Materials and Methods

The sparassodont from Cachapoal described here, SGOPV 3490, is preserved in highly indurated volcaniclastic matrix, as is the norm for mammal fossils from the Abanico Formation. Following mechanical preparation, the specimen was scanned at the PaleoCT facility at the University of Chicago using a µCT scanner (GE Phoenix v/tome/x 240kv/180kv scanner) to elucidate additional details of its morphology. The specimen was scanned at 180 kV and 150 µA with 0.5 mm Cu beam filter, producing a scan of 2021 slices with a voxel size of 0.058 mm. The specimen could not be segmented through automated thresholding due to extreme beam hardening and poor differentiation between rock and bone. As a result, segmentation was completed manually in Amira 5.3.3 (Visage Imaging Inc.). The specimen was visualized in Avizo 7.0.0 (Visualization Science Group) and Dragonfly 2.0 (Object Research Systems Inc.), and the latter was also used for measurements. Rotational movies of the segmented 3D model of this specimen are available as part of the online supplementary information (doi. org/10.5531/ sd.sp.41).

Data and measurements of other taxa are based on direct observation by R.K.E. or were taken from the primary literature. A complete list of specimens and references used for comparison is provided in appendix 1.

Saber teeth have evolved repeatedly among carnivorous mammals other than thylacosmiline sparassodonts, most famously in machairodontine felids but also in barbourofelid and nimravid carnivoramorphans as well as machaeroidine “creodonts” (which are probably members of Oxyaenidae; see Zack, 2019a). To avoid confusion, the term “sabertooth” is used here to refer to any carnivorous mammal with large, “saberlike” upper canines (saber teeth), whereas saber-toothed members of the Felidae are specifically referred to as “machairodontines.” The term “machairodonty” is used to refer to the saber-toothed condition and the associated morphological complex (Emerson and Radinsky, 1980; Antón, 2013), paralleling terminology used for other specialized craniodental morphologies (e.g., plagiaulacoidy, diprotodonty, hypselodonty).

Proborhyaenid (including thylacosmiline) sparassodonts are unusual among metatherians in having open-rooted, ever-growing (hypselodont) canines (Riggs, 1934; Simpson, 1948; Babot et al., 2002). However, some specimens suggest that canine roots closed and growth ceased in extreme senescence (i.e., MLP 79-XIII-18-1; see Bond and Pascual, 1983; Babot et al., 2002). Here, a tooth is regarded as hypselodont if it continued to grow and its roots remained open into the animal's adult lifespan. This definition mirrors the one used for notoungulates, in which taxa are considered hypselodont if their teeth are open rooted and ever-growing in adults but closed in extreme senescence (e.g., Adinotherium, Nesodon, and Trachytherus; Billet et al., 2008; Cassini et al., 2012, 2017).

The body mass of the new taxon described here was estimated using two methods: (1) lower molar row length (Lm1–4) using the dasyuromorphian regression equation of Myers (2001); and (2) length of m3, using the dasyurid regression equation of Gordon (2003). The first of these equations is the one most closely correlated with body size (i.e., it has the lowest percent prediction error) among the equations of Myers (2001) that can be applied to SGOPV 3490; the second was considered by Zimicz (2012) and Forasiepi et al. (2015) to be the best predictor of body mass in sparassodonts. Although the accuracy of these regression equations for estimating body mass in sparassodonts has been questioned due to extrapolation issues (most sparassodonts are far larger than any living carnivorous marsupial; Forasiepi et al., 2015), the taxon described here is small enough to cluster among the extant species used to construct the regression equation.

Phylogenetic Analysis

To establish the affinities of Eomakhaira, we carried out a phylogenetic analysis based primarily on the comprehensive matrix of Suarez et al. (2016), which is the most taxon-rich analysis of relationships within Sparassodonta to date. We augmented this base matrix with character and character-state data from Muizon et al. (2018). Features difficult to score in SGOPV 3490 because of poor preservation or substantial wear were coded as uncertain (i.e., “0/1”). The proborhyaenid Proborhyaena gigantea was added to the matrix to elucidate relationships within Proborhyaenidae; MLP 79-XII-18-1 (a specimen referred to P. gigantea by Bond and Pascual, 1983) was not included in this coding due to its uncertain taxonomic status (see Babot et al., 2002, and comments on this specimen below). Vincelestes neuquenianus was included in the analysis because TNT requires an a priori outgroup (Goloboff, 2009) and the branching sequence of plausible outgroups to Marsupialiformes (e.g., Holoclemensia, Deltatheroida, and Eutheria) is debated (see discussion in Beck, in press). A complete list of changes from Suarez et al. (2016) and Muizon et al. (2018), as well as the phylogenetic matrix used in this study in NEXUS format, is provided in the online supplementary information (doi. org/10.5531/sd.sp.41).

Dasyuromorphian taxa (Dasyurus, Sminthopsis, and Thylacinus) were constrained to form a monophyletic group in this study to avoid potential recovery of Sparassodonta within Dasyuromorphia, a problem encountered in previous studies (Forasiepi, 2009; Engelman and Croft, 2014; Forasiepi et al., 2015; Suarez et al., 2016). This problem is likely due to long branch attraction, limited outgroup sampling, and convergent evolution between Thylacinus and Sparassodonta. Most of the features shared between Thylacinus and Sparassodonta are apomorphies for Thylacinus within Dasyuromorphia (and thus do not characterize dasyuromorphians or thylacinids more broadly; Yates, 2014; Kealy and Beck, 2017; Rovinsky et al., 2019) and/or are associated with dietary habits (carnivory) and therefore potentially more homoplastic (Muizon and Lange-Badré, 1997).

The character-taxon matrix was compiled in Mesquite (Maddison and Maddison, 2008) and analyzed in TNT 1.1 (Goloboff et al., 2008) under equal weights, implied weights with a default concavity constant of k = 3, and implied weights with a higher concavity constant of k = 12, following suggestions that higher concavity constants produce more accurate results with larger datasets (Goloboff et al., 2018). Because the use of implied weighting in phylogenetics has been criticized (e.g., Congreve et al., 2016; Madzia and Cau, 2017; but see Goloboff et al., 2018), we used it primarily to evaluate support for hypotheses recovered using equal weights (i.e., topologies recovered using both methods were considered to be more robustly supported than topologies recovered via only one method). Tree analyses were performed in TNT using the “New Technology search” option, applying sectorial search, ratchet, tree drift, and tree fuse options under default parameters, finding the minimum length 1000 times and then searching within the set of recovered trees using tree bisection reconnection branch swapping.

Anatomical Abbreviations: Upper and lower incisors, canines, premolars, and molars are designated as I/i, C/c, P/p, and M/m.

Institutional Abbreviations: AC, Beneski Museum of Natural History, Amherst; AMNH, American Museum of Natural History, New York; CMNH, Cleveland Museum of Natural History, Cleveland; CORD-PZ, Museo de Paleontología, Facultad de Ciencias Exactas, Físicas y Naturales de la Universidad Nacional de Córdoba, Córdoba, Argentina; FMNH, the Field Museum, Chicago; IGM, Instituto Nacional de Investigaciones Geológico-Mineras, Bogotá, Colombia; MACN-A, Ameghino collection, Museo Argentino de Ciencias Naturales “Bernardino Rivadavia,” Buenos Aires, Argentina; MACN-PV, vertebrate paleontology collection, Museo Argentino de Ciencias Naturales “Bernardino Rivadavia,” Buenos Aires, Argentina; MHNT, Museu de História Natural de Taubaté, Taubaté, Brazil; MLP, Museo de La Plata, La Plata, Argentina; MMP, Museo Municipal de Ciencias Naturales de Mar del Plata, Mar del Plata, Argentina; MNHN, Muséum national d'Histoire naturelle, Paris, France; MNHN-Bol, Museo Nacional de Historia Natural, La Paz, Bolivia; MNRJ, Museu Nacional e Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil; MPEF-PV, Museo Paleontológico Egidio Feruglio, Trelew, Argentina; MUSM, Museo de Historia Natural de la Universidad Nacional Mayor San Marcos, Lima, Peru; PVL, Paleontología Vertebrados Lillo, Tucumán, Argentina; SGOPV, vertebrate paleontology collections, Museo Nacional de Historia Natural, Santiago, Chile; TMM, Texas Memorial Museum, Austin; UATF-V, Universidad Autónoma Tomás Frías, Potosí, Bolivia; UCMP, University of California Museum of Paleontology, Berkeley, California; UF, Florida Museum of Natural History, University of Florida, Gainesville; UNPSJB PV, vertebrate paleontology collection, Universidad Nacional de La Patagonia San Juan Bosco, Comodoro Rivadavia, Argentina; YPFB Pal, paleontology collection, Yacimientos Petrolíferos Fiscales de Bolivia in the Centro de Tecnología Petrolera, Santa Cruz, Bolivia; YPM-VPPU, Princeton University Collection, Yale Peabody Museum, New Haven.

Cranium

The maxilla of SGOPV 3490 is proportionally deeper than in most other sparassodonts, including the robust-skulled proborhyaenids Arminiheringia and Callistoe and borhyaenid Arctodictis sinclairi (fig. 5, table 1). Only Australohyaena antiquua, Arctodictis munizi, and Thylacosmilus atrox have relatively deeper maxillae among the taxa analyzed. A small portion of the dorsal border of the infraorbital foramen is preserved in SGOPV 3490 (fig. 3), indicating that this structure opened dorsal to the P3/M1 embrasure, as in Patagosmilus, Thylacosmilus, and Australohyaena but unlike in: (1) the Eocene taxa Callistoe and Arminiheringia, in which the infraorbital foramen opens above or anterior to the anterior root of P3; (2) Borhyaena and cf. Proborhyaena (MLP 79-XIII-18-1), in which the foramen opens slightly more anteriorly over the posterior root of P3; and (3) species of Arctodictis, in which the foramen is more posterior (above or posterior to the posterior root of M1; Forasiepi, 2009).

TABLE 1.

Measurements of the holotype of Eomakhaira molossus (SGOPV 3490) in mm. Greatest dorsoventral height of maxilla measured from alveolar border of P3 to dorsal border of maxilla. Although the outer portion of the maxilla is damaged at this level (see fig. 3), enough of the medial surface of the maxilla and premolar alveoli is preserved to be able to determine that P1–3 are in life position.

Although the alveolar border of the maxilla posterior to the infraorbital foramen is fragmentary, the parts preserved suggest that Eomakhaira lacked maxillary “cheeks,” i.e., protrusions of the maxilla posterior to the infraorbital foramen that extend lateral to the toothrow (best seen in ventral view). Maxillary “cheeks” are a highly variable feature within Sparassodonta. They are present in borhyaenids, Prothylacynus, and many hathliacynids but absent in most species of Lycopsis (except L. torresi), Acyon, Patagosmilus, and cf. Proborhyaena (MLP 79-XII-18-1). Contrary to some reports, maxillary “cheeks” appear to be absent in Paraborhyaena and Thylacosmilus (FMNH P14531, MLP 35-X-4-1, MMP 1443; see also Petter and Hoffstetter, 1983; Goin and Pascual, 1987). The state in Arminiheringia could not be determined based on available information.

Based on the posterior border of the left nasal (fig. S1), the two naso-frontal sutures of Eomakhaira form an angle of ∼100° in dorsal view, greater than the acute-angled naso-frontal sutures of Callistoe, but narrower than those of Paraborhyaena, Patagosmilus, Pharsophorus, and borhyaenids. An internasal projection of the frontals is absent (i.e., the naso-frontal suture is V-shaped rather than W-shaped). The preserved lateral edge of the left nasal is straight, suggesting it represents the border of the naso-lacrimal suture, based on a similar morphology of this suture in other sparassodonts. As in Callistoe, a portion of the nasals may have extended onto the lateral surface of the snout, but this cannot be determined with certainty. The nasals are proportionally slender compared to the rest of the skull. Scaling the greatest width of the nasals to the length of M3, the nasals of Eomakhaira are narrower than in most borhyaenoids except Callistoe, Patagosmilus, and a juvenile specimen of Prothylacynus patagonicus (MACN-A 5931), and they are much narrower than the nasals of Paraborhyaena, Arminiheringia, Patagosmilus, and borhyaenids (table S2). The nasals of SGOPV 3490 are only about 25% as wide anteriorly as they are at their widest point. In borhyaenoids, this figure is typically ∼30% (table S2), with the exception of Callistoe, in which the nasals vary less in width along their length (though this may be affected by mediolateral compression of the holotype). Nasal bones of the hathliacynids Sipalocyon and Acyon show less anterior tapering than in borhyaenoids apart from Callistoe, while the proportions of the nasals in Cladosictis more closely resemble those of borhyaenoids. Interestingly, the sparassodont UF 27881 from the middle Miocene of Quebrada Honda, Bolivia, originally described as a basal sparassodont (Engelman and Croft, 2014) but recovered as a borhyaenoid in later analyses (Forasiepi et al., 2015; Suarez et al., 2016) does not resemble borhyaenoids in its nasal proportions; rather, it is more similar to a specimen tentatively assigned to the basal sparassodont Hondadelphys (IGM 250364; Goin, 1997).

FIG 5.

Relative maxilla height and dentary depth (measured at m3–4 embrasure) in sparassodont specimens for which both maxilla and dentary are known, scaled to lower molar row length. Eomakhaira molossus is denoted by a star. Skulls of several taxa are illustrated to highlight variation. Skulls of Arctodictis sinclairi, Acyon myctoderos, Callistoe vincei, and Thylacosmilus atrox modified from Forasiepi (2009), Forasiepi et al. (2006), Babot et al. (2002), and Riggs (1934), respectively. Data for this figure can be found in table S1.

FIG 6.

Cranium of the holotype of Eomakhaira molossus (SGOPV 3490) in palatal view. Anterior to right. Colors of elements in this CT segmentation are the same as in figure 3. Abbreviations: mpps, medial postpalatine spine; pp, palatal pit; ptor, palatine torus; dental abbreviations as in Materials and Methods. Scale = 30 mm.

The palate of SGOPV 3490 is fragmented and patchily preserved, but enough is present to establish that Eomakhaira lacked maxillopalatine fenestrae, as in other sparassodonts (fig. 6). The length/width ratio of the palatal process of the maxilla exceeds 1.5, even taking distortion into account, a value similar to most sparassodonts (with the exception of thylacosmilines and the borhyaenids Australohyaena and Arctodictis, in which the process is anteroposteriorly shorter and mediolaterally wider and the ratio is less than 1.5). A pair of palatal pits occurs on the palatal process of the maxilla between M3–4; whether an additional pair was present between M2–3 cannot be determined. Accounting for deformation, anterior displacement of the palatine, and the separation of the maxillary and palatine borders of the minor palatine foramen (see below), the horizontal process of the palatine does not appear to have extended posterior to M4, reminiscent of the condition in Borhyaena, Patagosmilus, and some specimens of Prothylacynus.

A pair of low palatine tori are present at the posterior end of the palate. These structures are mediolaterally broad, extending across each palatine bone, but do not contact one another medially (fig. 7). Among metatherians, sparassodonts are unusual in the general absence of a palatine torus, a feature also observed in deltatheroidans (Forasiepi, 2009; Bi et al., 2015), basal didelphids (caluromyines and Glironia; Voss and Jansa, 2009), and some dasyuromorphians (i.e., Thylacinus; Wroe, 1999; Warburton et al., 2019; R.K.E., personal obs.). In all of these taxa, the posterior border of the palate is typically single or double arched in ventral view (depending on the presence or absence of a medial postpalatine spine). Although most sparassodonts lack a palatine torus, in many (e.g., UF 27881, Cladosictis, Sipalocyon, Borhyaena, Australohyaena, Arctodictis, Thylacosmilus) the posterior border of the palate is slightly thickened. This thickening can be extremely pronounced (e.g., as in species of Arctodictis; Forasiepi et al., 2004; Forasiepi, 2009), but it follows the borders of the choanae rather than forming a straight torus. Other than Eomakhaira, the only sparassodonts with true palatine tori are the basal taxon Allqokirus and the proborhyaenids Callistoe and Arminiheringia. However, even in these taxa, the palatine tori do not resemble those of most other metatherians, wherein a straight palatine torus defines the posterior border of the palate. In Allqokirus, the palatine torus is well developed but is posteriorly concave rather than straight. Callistoe and Arminiheringia resemble Eomakhaira in having palatine tori that are mediolaterally oriented ridges at the back of the palate. Compared to the thickened choanal border of other sparassodonts, these structures are much more prominent in Eomakhaira, forming distinct processes that do not follow the borders of the choanae. However, unlike in other metatherians, the palatine tori in Callistoe and Arminiheringia do not form a single, complete torus; rather, they are paired structures that do not meet at the midline (except possibly in Callistoe) and do not constrict the choanae (i.e., in ventral view, the choanae still exhibit the classic “double arch” pattern typical of sparassodonts).

FIG 7.

Posterior palate of the holotype of Eomakhaira molossus (SGOPV 3490), in oblique anterior view. Anterior to lower left. Shows the paired palatine tori and broken border of the minor palatine foramen. Colors of elements in this CT segmentation are the same as in figure 3. Abbreviations: mpf, minor palatine foramen; ptor, palatine tori. Scale = 10 mm.

A small foramen occurs just lateral and slightly dorsal to the lateral edge of each palatine torus. It likely represents the minor palatine foramen (fig. 7) based on its position, though it is slightly damaged on both sides of the skull due to anterior displacement of the palatines. In Eomakhaira, the minor palatine foramen is located between the maxilla and the palatine, as in most metatherians, including many sparassodonts (including UF 27881, Cladosictis, Arctodictis, Callistoe, Arminiheringia, Patagosmilus, and some but not all specimens of Thylacosmilus). In Callistoe and Arminiheringia, as in Eomakhaira, the minor palatine foramen is dorsal to and slightly tucked under the lateral edge of the palatine torus. The minor palatine foramen is positioned more laterally in Eomakhaira than in other sparassodonts; in most sparassodonts, it is closer to the choanae than to the upper teeth, whereas in Eomakhaira, it is closer to the upper dentition. In this respect, Eomakhaira resembles Patagosmilus (but not Thylacosmilus). The minor palatine foramen of Eomakhaira is fairly large, more comparable in size to that of Patagosmilus than Arminiheringia or Callistoe (in which it is smaller). This foramen is clearly not the postpalatine torus foramen present in most groups of New World metatherians (Wible, 2003), as the homologous structure is either an open notch or absent in sparassodonts (Muizon et al., 2018). The postpalatine torus foramen opens directly posteriorly into the basipharyngeal canal in other metatherians, whereas the inferred minor palatine foramen of Eomakhaira opens posterolaterally into the orbitotemporal region (fig. 8), similar to the path of the minor palatine foramen of other sparassodonts. The postpalatine torus foramen appears to be absent in Eomakhaira, as it is in proborhyaenids (including thylacosmilines) and borhyaenids (Muizon et al., 2018).

SGOPV 3490 preserves a small portion of the orbital region, primarily on the left side. This consists of part of the ascending process of the palatine posteriorly and several isolated plates of bone separated by distinct gaps and holes that form part of the orbital wall anteriorly (fig. 8). Based on their position and morphology, these bone fragments likely represent parts of the orbital process of the lacrimal and the anterior part of the ascending process of the palatine. No clear suture is visible between the palatine and maxilla in lateral view. Based on size and location, these gaps may represent sutures and foramina that were enlarged postmortem by fragmentation of the fragile surrounding bone. The largest of these openings compares well to the sphenopalatine foramen in location. This foramen opens approximately dorsal to M4, as it does in some other borhyaenoids (borhyaenids, Callistoe, Thylacosmilus).

The anteriormost fragment of the orbital wall appears to represent a small portion of the orbital process of the lacrimal. Parts of the facial and zygomatic processes of the lacrimal may be present but cannot be distinguished from surrounding elements (i.e., maxilla). At the anterior end of the fragment of the orbital process is a small, partially preserved canal, likely the lacrimal foramen. Based on its position and surrounding elements, the lacrimal foramen appears to have opened within the orbit. Most sparassodonts have a single lacrimal foramen opening inside the orbit. However, in Mayulestes, Allqokirus, Lycopsis padillai, and Arminiheringia there are two lacrimal foramina, and in Callistoe, the lacrimal foramen number is polymorphic (Suarez et al., 2016; Muizon et al., 2018). Additionally, in Allqokirus and Mayulestes, one of the two lacrimal foramina is laterally exposed rather than enclosed within the orbit (Muizon et al., 2018). It is clear that at least one lacrimal foramen that opened within the orbit was present in SGOPV 3490, though the lacrimal foramen count is uncertain due to the limited preservation of this element.

FIG 8.

Left orbital region of the holotype of Eomakhaira molossus (SGOPV 3490) in oblique posterolateral view. Anterior to left. Colors of elements in this CT segmentation are the same as in figure 3. Abbreviations: laf, lacrimal foramen, mpf, minor palatine foramen; spf, sphenopalatine foramen. Scale = 30 mm.

SGOPV 3490 preserves parts of both jugals. On the left side, this element is represented by a fragmentary bone “floating” near the orbital region. Part of the right jugal also seems to be present, represented by small patches of bone (including parts of the rostrum formed by the jugal in other metatherians) and remnants of a marrow cavity dorsal to the upper molars. The shape and location of the maxillo-jugal suture cannot be determined.

Mandible

The mandible of Eomakhaira is robust and deep (fig. 9). It is comparatively shallower than the mandible of the proborhyaenids Callistoe and Arminiheringia (fig. 5, tables S1, S16) and deeper than in Proborhyaena and Paraborhyaena. The dentary is fractured parallel to the long axis of the horizontal ramus of the dentary, and these complementary fractures are displaced and infilled by matrix (fig. 9C, D). The thickness of these matrix-infilled gaps between complementary bone fragments suggests that the dentary of SGOPV 3490 was originally ∼2+ mm shallower than what is reported here. As in Callistoe, Arminiheringia, and Pharsophorus tenax, the horizontal ramus of Eomakhaira deepens posteriorly, has a ventral border that is curved in lateral view, and terminates anteriorly in a simple curve. Indeed, the horizontal ramus in Eomakhaira strongly resembles that of Arminiheringia (MACN-A 10970). This contrasts with the condition in the borhyaenids Australohyaena and Arctodictis, the proborhyaenids Proborhyaena and Paraborhyaena, and MPEF-PV 4170, a specimen assigned to Pharsophorus cf. P. lacerans (but not in the holotype of P. lacerans, MACN-A 52-391), in which the horizontal ramus is nearly uniform in depth, its ventral margin is flat, and its anterior end forms a distinct “chin” (fig. S2), with a sharp angle between the anterior border of the symphysis and the ventral border of the horizontal ramus. A distinct “chin” is also present in the basal sparassodonts Allqokirus and Mayulestes but is absent in most other members of this group in which the anterior border of the dentary is curved. The dentary shows no sign of a genial flange, in contrast to Thylacosmilus, Anachlysictis, or the unnamed thylacosmilid from La Venta (in which it is present). The deepest point of the horizontal ramus appears to have been below m4 in SGOPV 3490, as in most sparassodonts (table 1).

The left dentary is displaced slightly anteriorly relative to the right. The anteroventromedial edge of the left dentary is straight in ventral view (fig. S3), and the medial face of this element preserves a small portion of the symphyseal surface (fig. 9D). The symphyseal surface does not bear well-developed interdigitating ridges (in contrast to sparassodonts like Borhyaena, in which such ridges are present). In dorsal and ventral views, the medial face of the right dentary inflects medially near the p2/3 embrasure, a feature denoting the posteriormost extent of the mandibular symphysis in most mammals, suggesting the symphysis of Eomakhaira extended posteriorly to the level of the p2/p3 embrasure, or at most, just slightly below the anterior root of p3 (fig. 9). The symphysis of Eomakhaira is thus rather short compared to closely related sparassodonts; in these taxa, the mandibular symphysis reaches its midpoint below the: (1) p3 roots (Borhyaena macrodonta, Borhyaena tuberata, and Pharsophorus lacerans [including MPEF-PV 4190, the specimen of Pharsophorus cf. P. lacerans from La Cantera]), (2) posterior root of p3 (Plesiofelis, Australohyaena and Prothylacynus), (3) p3/m1 embrasure (Callistoe, Paraborhyaena, Proborhyaena, and Arctodictis spp.), or (4) m1 (Arminiheringia; Babot et al., 2002; Zimicz, 2012). In Acrocyon riggsi (Goin et al., 2007) and Pharsophorus tenax, the symphysis extends below the anterior root of p3 but further posteriorly than in SGOPV 3490. In Thylacosmilus atrox and Anachlysictis gracilis, the symphysis is much shorter than in all aforementioned taxa, ending below the canines.

FIG 9.

Mandible of the holotype of Eomakhaira molossus (SGOPV 3490). Right dentary in A, labial and C, lingual views. Left dentary in B, labial and D, lingual views. Colors of elements are the same as in figure 3. Abbreviations: cor, coronoid process of dentary; sulc, lingual sulcus of the lower canine; menf, mental foramina; symph, mandibular symphysis. Scale = 50 mm.

Several lines of evidence suggest that the mandibular symphysis of SGOPV 3490 was unfused. The two dentaries are offset anteroposteriorly with respect to one another. The preserved portion of the left symphyseal surface bears a straight ventromedial edge, suggesting a clean break between the two rami, as would be expected if they had been held together by ligaments. This configuration would be unlikely if a fused symphysis were broken postmortem. Had the mandibular symphysis been fused in vivo, one would expect an uneven, jagged break between the two dentaries, or for one or both to be broken immediately posterior to the mandibular symphysis, where the dentary is comparatively the weakest. This is the case in several other sparassodonts with fused symphyses and broken mandibles (e.g., MACN-A 706, Prothylacynus patagonicus; MLP 85-VII-3-1, Arctodictis sinclairi; UATF-V-000129, Paraborhyaena boliviana).

Each dentary of SGOPV 3490 bears only two mental foramina, one located beneath p2–3 and another below the m1–2 embrasure. The posterior foramen is well defined on both dentaries and opens posteriorly; the anterior foramen is not well preserved on either dentary, but its existence and position can be inferred from a gap in the bone fragments labially beneath the premolar row. The presence of only two mental foramina on each dentary in Eomakhaira is unusual for a borhyaenoid. Most hathliacynids, borhyaenoids, and the basal sparassodont Stylocynus bear three or more mental foramina on each dentary, and individuals of some species possess as many as five or six (Acrocyon riggsi, Arctodictis sinclairi, Australohyaena antiquua, Borhyaena tuberata, and Lycopsis longirostrus). Additional mental foramina may have been present in SGOPV 3490, but the location of the preserved bone fragments relative to the positions of mental foramina in other sparassodonts (typically between p2 and m1 or m2) make this unlikely.

Enough of the left coronoid process is preserved to indicate that it is tall and well developed. However, due to crushing and distortion, it is not possible to determine the shape of the masseteric fossa, nor can the angle between the anterior border of the coronoid process and the toothrow be securely determined. A small portion of the right ascending ramus (which is not obviously deformed) indicates that the anterior border of the coronoid process is oriented approximately 110°–113° relative to the toothrow, a value typical for sparassodonts (Forasiepi, 2009).

Dentition

Precanine Dentition: Little of the precanine dentition is preserved in SGOPV 3490. A cylindrical fragment of a small tooth appressed to the lingual side of the lower right canine may represent i3, based on the position of this tooth in other sparassodonts (fig. 15). If this fragment represents part of a lower incisor, then the lower incisors of Eomakhaira molossus were proportionally smaller than those of Arminiheringia auceta, Arctodictis sinclairi, and potentially even Australohyaena antiquua and Paraborhyaena boliviana (scaling by both p3 and m4) but still larger than in MLP 77-VI-13-1, a specimen assigned to Arctodictis sinclairi that has been noted to have relatively small teeth compared with other specimens of this taxon (Goin et al., 2007).

Canines: The most conspicuous feature of the holotype of Eomakhaira is its large, robust canines (fig. 3, table 2). These teeth are disproportionately large compared to most sparassodonts, comparable (in relative size) only to proborhyaenids (including thylacosmilines), and the borhyaenids Australohyaena, Acrocyon, and Arctodictis (table 3). The surfaces of the canine roots in SGOPV 3490 are smooth (fig. 4A, 10A). In most borhyaenoids, the canine roots bear a series of small longitudinal grooves that sometimes nearly reach the apex (e.g., in Arminiheringia and Proborhyaena; figure S4B). The only borhyaenoids that do not have these grooves are thylacosmilines (figure S4C) and possibly Lycopsis viverensis (Suarez et al., 2016).

The upper and lower canines of Eomakhaira bear a well-developed median sulcus lingually, making them somewhat reniform in cross section (fig. 10B, C). By contrast, no sulcus occurs on the labial side of lower canine or the exposed labial surface of the upper canine. A shallow labial sulcus is present on intralveolar portions of the upper canine but is visible only on CT scans. Although median canine sulci have been considered a synapomorphy of proborhyaenids (either sensu stricto or including thylacosmilines; Babot et al., 2002), this feature is widely distributed among sparassodonts, with lingual sulci also occurring in the borhyaenids Australohyaena antiquua, Arctodictis sinclairi, Arctodictis munizi, and Borhyaena macrodonta, the basal borhyaenoid Pharsophorus lacerans (as seen in the holotype MACN-A 32-391, YPM-VPPU 20551, and MPEF-PV 4190, a specimen from La Cantera assigned to Pharsophorus cf. P. lacerans; Patterson and Marshall, 1978; Goin et al., 2010), and an indeterminate sparassodont from the Fitzcarrald Arch (Tejada-Lara et al., 2015). Sinclair (1930) and Marshall (1978) reported lingual sulci in Acrocyon riggsi, but we could not verify this observation in photographs of this taxon. In all sparassodonts in which median sulci occur, the labial ones are less prominent than the lingual ones (e.g., proborhyaenids, Arctodictis munizi) or are absent (all other taxa). Although the extraalveolar portion of the upper canine of Thylacosmilus lacks a median sulcus, a shallow median sulcus is present on the intralveolar portion of this tooth (FMNH P14344, FMNH P14531). The lower canines of Thylacosmilus bear median sulci labially and lingually (Goin and Pascual, 1987).

TABLE 2.

Dental measurements of the holotype of Eomakhaira molossus (SGOPV 3490) in mm. Measurements from CT scans, to nearest 0.1 mm. * = estimated measurement due to damaged protocone and missing metastylar corner of M3.

In cross section, the lingual side of the upper canines of SGOPV 3490 is slightly flatter than the labial side but not as flat as in Thylacosmilus atrox (Riggs, 1934). The labial surface of the right upper canine of Eomakhaira is slightly keeled (fig. 4A, 10B, C), similar to but less pronounced than in Thylacosmilus (Riggs, 1934; Turnbull, 1978). This feature is not evident on the more poorly preserved left upper canine. A small carina occurs on the posterior edge of the upper canine, slightly more marked on the right tooth than the left (fig. 10B). Carinae appear to be absent in the upper canines of Arminiheringia auceta, cf. Proborhyaena (MLP 79-XII-18-1), and other borhyaenids we have observed. By contrast, the upper canines of thylacosmilines, like those of most saber-toothed mammals, bear well-defined carinae that lend the tooth a knifelike appearance. In Patagosmilus, the upper canine is blunt anteriorly but bears a sharp posterior carina (Forasiepi and Carlini, 2010), similar to the condition in Eomakhaira. The anterior and posterior faces of the upper canines form well-defined carinae in Thylacosmilus, the posterior of which is much sharper (Riggs, 1934; Turnbull, 1978; Goin and Pascual, 1987).

FIG 10.

A, Photograph and B, C, CT images of the right upper canine of the holotype of Eomakhaira molossus (SGOPV 3490). A, Lateral view, showing bluntness of canine apex and the absence of enamel, longitudinal ridges, and labial median canine sulcus. B, Transverse section of canine, slightly below alveolar border (actual point at alveolar border obscured by a crack), showing posterior keel. Anterior root of P1 (aP1) marked to show that the longitudinal ridge is not an artifact of postmortem damage. C, Transverse section of canine, at level of tooth row, showing presence of median labial keel and lingual median sulcus. Approximate location of sections in B, C denoted by arrows on A. Anterior to right in all images, and lingual to top in B, C. Abbreviations: aP1, anterior root of P1; mk, labial median keel; ms, lingual median sulcus; pr, posterior ridge. Scale = 10 mm (A); 5 mm (B, C).

The upper canines of Eomakhaira are mediolaterally compressed compared to non-thylacosmiline borhyaenoids (table 3). The L/W ratio of the upper canines ranges from 1.50–1.77, depending on orientation of the skull during measurement. The lower end of this range likely underestimates the degree of mediolateral compression of the tooth, however, as it requires an anatomically unlikely orientation of the specimen (one where the palate is highly inclined and the tooth roots are far from vertical). More reasonable orientations of the specimen yield higher estimates. The most reasonable orientations of SGOPV 3490 produce canine L/W ratios of 1.65–1.70. Orientation aside, the upper canines of Eomakhaira are clearly more mediolaterally compressed than in most other borhyaenoids, including Pharsophorus, Prothylacynus, all borhyaenids (Acrocyon, Arctodictis, Australohyaena, and Borhyaena), IGM 251108 (the putative thylacosmiline from La Venta), most non-thylacosmiline proborhyaenids (Arminiheringia, Paraborhyaena, Callistoe), and the indeterminate proborhyaenid from the Tremembé Formation (Couto-Ribeiro, 2010). However, they are less compressed than in Patagosmilus, Thylacosmilus, and cf. Proborhyaena (MLP 79-XII-18-1). In terms of nonsparassodont carnivores, the canine proportions of Eomakhaira are comparable to the machaeroidine oxyaenid “creodont” Machaeroides eothen (see Gazin, 1946), the nimravid carnivoramorphans Dinictis felina and Nimravus brachyops (see Barrett, 2016), and the machairodontine felid carnivoran Pseudaelurus quadridentatus (see Antón et al., 2012). The resemblance to the latter three placental taxa is noteworthy, as each is among the least-specialized members of their respective clades in terms of machairodonty (Meachen-Samuels, 2012; Antón, 2013).

TABLE 3.

Upper canine cross-sectional shape (ratio of anteroposterior length [L] to mediolateral width [W]) and relative canine size of SGOPV 3490 compared with other sparassodonts. * = juvenile individuals. For raw measurements and relative canine size calculation method, see table S3.

continued

CT imaging indicates that the canine roots of SGOPV 3490 were closed at the time of death. The canine roots are closed in adulthood in sparassodonts except in non-thylacosmiline proborhyaenids, in which the upper and lower canines are hypselodont and their roots remain open throughout life (Simpson, 1948; Marshall, 1978; Babot et al., 2002; but see Bond and Pascual, 1983). In thylacosmilines, only the upper canines are hypselodont (Riggs, 1934; Forasiepi and Carlini, 2010). However, the closed roots of the lower canines may be a secondary reversal from a hypselodont ancestral condition (Babot et al., 2002). Canine pulp cavities in SGOPV 3490 are relatively narrow and lack a well-developed opening. Additionally, the canine roots taper slightly apically rather than being uniformly wide with outwardly flaring edges as is typically observed in open-rooted taxa. Nevertheless, several features suggest that the canines of Eomakhaira were open rooted earlier in ontogeny, closing only in extreme senescence, as in the proborhyaenid Proborhyaena (Bond and Pascual, 1983; but see Babot et al., 2002). First, the canines of SGOPV 3490 lack any trace of enamel, being composed solely of dentine. This contrasts with the typical condition in sparassodonts with nonhypselodont canines (e.g., borhyaenids), where some enamel occurs on nonoccluding surfaces near the apex of the tooth (such as the lateral side of the upper canines), even in old individuals. The hypselodont canines of Patagosmilus and Thylacosmilus have enamel (Turnbull, 1978; Forasiepi and Carlini, 2010), but only in a band covering the labial surface of these teeth rather than a simple cap. Postmortem damage and/or unusual wear are unlikely to account for the complete absence of enamel on all four canines of SGOPV 3940, particularly considering that enamel occurs on the right P1 and p2, located only a few millimeters posterior to the canines (fig. 4). A similar condition (enamel absent on the canines but present on the postcanines) is also observed in the proborhyaenids Proborhyaena and Callistoe. The absence of enamel in Eomakhaira is consistent with this taxon having had open-rooted canines until near the end of its lifespan.

The upper canines of SGOPV 3490 are tall given their high degree of wear (table S4). The upper canine roots extend almost to the dorsal border of the maxilla within their alveoli (measuring over 46 mm in total length), indicating that the height of the exposed portion (23 mm) is natural and not the result of the canine slipping ventrally from the alveolus. Scaled to the anteroposterior length of the tooth, the exposed height of the upper canine in SGOPV 3490 exceeds that of most other borhyaenoids and far exceeds that of other sparassodonts with a comparable degree of canine wear. Scaled to M3 length, the upper canine is longer than in almost any other taxon, except those with hypselodont upper canines.

The pulp cavities of the upper and lower canines appear to extend to their apices. The cavity is difficult to discern in the right upper canine, but an area of seemingly less dense, nodule-containing material runs the entire length of the tooth. The pulp cavities of the upper canines are much smaller than those of the lower canines. A pulp cavity extending to the apex of the tooth would be expected if these teeth were worn but hypselodont (O'Connor et al., 2019), as is seen in other mammals with hypselodont caniniforms (i.e., Choloepus; DigiMorph Staff, 2003), though this condition can also result from extreme wear in some sparassodonts without hypselodont canines (e.g., the holotype of Pharsophorus lacerans, MACN-A 52-391).

The left lower canine, which has the best-preserved root among the four canines, may not have a fully closed root. Although its pulp cavity is very narrow along most of its length, a deep depression occurs near its base, where the edges of the root flare outward in cross section (fig. 11). This depression connects with the pulp cavity of the tooth. This is reminiscent of the condition in non-thylacosmiline proborhyaenids, all of which have hypselodont lower canines (Babot et al., 2002: fig. 4B, E). The lack of a basal tip on the left lower canine of SGOPV 3490 cannot be ascribed to damage, as the base of the alveolus is intact and the edges of the apical depression are smooth and rounded rather than jagged, as they would be if broken. Given the advanced ontogenetic age of SGOPV 3490, details of the root of the lower left canine suggest that it had been open for much of the animal's life and had only very recently closed. The conditions of the roots of the upper canines are uncertain, as their bases cannot be confidently distinguished from the surrounding matrix and bone. Together, the observations above suggest that the canines of Eomakhaira were hypselodont throughout most of their ontogeny, with roots closing and tooth growth ceasing only in extremely senescent individuals.

FIG 11.

Morphology of the left lower canine root in the holotype of Eomakhaira molossus (SGOPV 3490). A, CT reconstruction in oblique lateral view, showing prominent depression on proximal end of canine and its connection to the pulp cavity. Connection to pulp cavity has been artificially darkened for contrast. B, Oblique lateral CT image slice of SGOPV 3490, showing flared, smooth distal end of left lower canine (denoted by arrow). Scale = 10 mm.

The roots of the canines in Eomakhaira, particularly the lower canines, are shorter and less curved than those of other sparassodonts. The upper canine roots end above P3 in Eomakhaira but are still relatively large above P3 in Callistoe and Arminiheringia (based on coronal cross sections; Babot et al., 2002), suggesting that the roots extended further posteriorly in those taxa. A condition similar to that of Callistoe and Arminiheringia occurs in Australohyaena (UNPSJB PV 113; Forasiepi et al., 2015). The lower canines of SGOPV 3490 are emplaced nearly subvertically within the alveoli, similar to what has been described for most borhyaenoids, with the notable exception of Arminiheringia, in which the lower canines are procumbent. The lower canine roots are much shorter and more vertical than their upper counterparts and probably end at the level of the p1–2 embrasure. However, they certainly do not extend beyond the anterior root of p2. This differs from Callistoe, Proborhyaena, and Arctodictis, in which the lower canine roots reach the level of p3 (Babot et al., 2002; Forasiepi, 2009), and from Arminiheringia, in which they reach the molar row (Zimicz, 2012). The lower canine roots are more curved in Callistoe and Arctodictis than in Eomakhaira.

Premolars: SGOPV 3490 possesses the typical metatherian postcanine dental formula of three premolars and four molars (fig. 15). In the dentary, the alveolar margin of the postcanine dentition is higher lingually than labially, a feature noted in other sparassodonts (Forasiepi et al., 2015). The postcanine toothrow of Eomakhaira is straight in occlusal view (fig. 6), as in most sparassodonts. In the thylacosmilines Patagosmilus and Thylacosmilus, the postcanine toothrow is strongly curved and concave medially (Forasiepi and Carlini, 2010), but this does not appear to be the case for the more basal thylacosmiline Anachlysictis (Goin, 1997). In Arctodictis and Australohyaena, the postcanine toothrow is neither straight nor curved. Instead, the long axes of the molar and premolar rows are offset from one another rather than aligned (fig. S5), a feature often associated with an obliquely oriented P3. This condition is more pronounced in Arctodictis munizi and Australohyaena antiquua than in Arctodictis sinclairi. The condition in the borhyaenid Acrocyon riggsi is unclear; in the holotype (FMNH P13433), the toothrow is straight, but P3 is slightly oblique. This unusual configuration may be an artifact of poor restoration of the holotype skull (Sinclair, 1930). FMNH P13433 is also unusual in having postcanine toothrows that do not diverge posteriorly to form a triangular palate. Posteriorly diverging toothrows characterize every sparassodont for which the shape of the palate can be determined except Hondadelphys (thought be a basal member of the clade), an observation that supports the interpretation of Sinclair (1930) that the straight postcanine toothrows of FMNH P13433 are an artifact. The long axis of P3 in Eomakhaira is parallel to the long axis of the postcanine toothrow (fig. 6), as in Pharsophorus, Borhyaena, and proborhyaenids, rather than being obliquely oriented as in Australohyaena and Arctodictis.

As in other sparassodonts, the upper and lower premolars of Eomakhaira increase in size from P1 to P3 and p1 to p3. However, the relative disparity in sizes among the premolars differs between the upper and lower toothrows. P3 is much larger than P1and P2, which are of similar size. By contrast, p1 is distinctly smaller than p2 and p3, which are of similar size. The first upper and lower premolars are both very small. In lateral view, the dorsal alveolar border of the lower premolars slopes anterodorsally-posteroventrally from p1 to p3, a common pattern also seen in Prothylacynus patagonicus, Pharsophorus cf. P. lacerans (MPEF-PV 4170), Borhyaena macrodonta, Arctodictis sinclairi, Australohyaena antiquua, Callistoe vincei, and Proborhyaena gigantea. The premolar row is relatively short in Eomakhaira compared to other sparassodonts (table S5).

The right P1 and both p1s are oriented obliquely relative to the remainder of the toothrow (∼35° anterolabially-posterolingually for both loci), but the left P1 is nearly parallel to the toothrow. Given that the left maxilla is poorly preserved, the anterior root of its P1 is damaged, and natural bilateral asymmetry of tooth orientation has never been documented in Sparassodonta, the orientation of this tooth almost certainly reflects postmortem deformation.

The P1 of Eomakhaira is asymmetric in lateral view, with its main cusp located over the anterior root (fig. 4) rather than equidistant between the anterior and posterior roots, as in most sparassodonts. Similarly asymmetric premolars occur in Allqokirus and Mayulestes but not in hathliacynids, borhyaenids, or most basal borhyaenoids (Muizon et al., 2018). They do occur in the proborhyaenids Callistoe and Arminiheringia, the only other proborhyaenids for which P1 is known. P/p1 is thought to have been lost in thylacosmilines (Forasiepi and Carlini, 2010), but the putative basal thylacosmiline from La Venta retains P/p1 (as evidenced by the presence of the roots of these teeth in IGM 251108). What remains of p1 in SGOPV 3490 suggests that its apex was more centrally positioned than that of P1, as in other sparassodonts.

The crowns and roots of P/p2 of Eomakhaira are aligned with the rest of the postcanine toothrow, as in Pharsophorus (though possibly not as in MPEF-PV 4170 from La Cantera, which more closely resembles Arctodictis sinclairi in this respect), Callistoe, Arminiheringia auceta, and Borhyaena, rather than oblique to the toothrow, as in Proborhyaena, Paraborhyaena, Arminiheringia contigua (MACN-A 10317), Arctodictis, Acrocyon, and Australohyaena.

The robust P/p3 bear large, stout roots. Nevertheless, these teeth are proportionally narrower labiolingually than the more bulbous teeth of borhyaenids (P3 L/W ratio = 1.75 in Eomakhaira vs. an average of 1.43 in borhyaenids; table S6) and those of species of Pharsophorus (P. lacerans, YPM-VPPU 20551; P. tenax, AC 3192). Borhyaenid taxa closely resembling Eomakhaira in other respects (e.g., Arctodictis and Australohyaena) also have the most bulbous P/p3s. The P3 of Eomakhaira is more elongate than that of Callistoe vincei, less elongate than that of cf. Proborhyaena (MLP 79-XII-18-1), but comparable to those of specimens of Arminiheringia. Like P3, the p3 of Eomakhaira is proportionally narrower labiolingually than in most borhyaenids, the basal borhyaenoid Plesiofelis, and the proborhyaenids Proborhyaena and Arminiheringia. The L/W ratio of this tooth is comparable to that in the proborhyaenids Paraborhyaena and Callistoe, but is less narrow than its counterparts in the basal borhyaenoids Prothylacynus and Pharsophorus.

The P3 of Eomakhaira is ∼13%–19% longer than p3. In most sparassodonts resembling Eomakhaira (borhyaenids, proborhyaenids, and Pharsophorus), P3 and p3 are of similar length (table S6). The only borhyaenoid potentially resembling Eomakhaira in this respect is Proborhyaena gigantea. The P3 of MLP 79-XII-18-1, assigned to Proborhyaena by Bond and Pascual (1983) but referred to as cf. Proborhyaena here, is extremely large (nearly 30 mm long, judging from its preserved roots). This is considerably longer than the p3 of the largest known specimen of Proborhyaena gigantea (AMNH 29576, where this tooth is ∼24 mm long). If MLP 79-XII-18-1 pertains to Proborhyaena, it implies the P3 of this taxon was >50% longer than p3, a more extreme size disparity than in Eomakhaira. The tooth at the P3 locus in Thylacosmilus is also much longer than its p3, but since the upper tooth represents dP3 rather than P3 (Goin and Pascual, 1987; Forasiepi and Carlini, 2010; Forasiepi and Sánchez-Villagra, 2014), direct comparisons with other sparassodonts are not possible. No other thylacosmilines are known from associated upper and lower dentitions.

The p3 of Eomakhaira appears to have been inclined posteriorly. This is common among borhyaenoids, occurring also in the basal forms Plesiofelis schlosseri and Pharsophorus lacerans; the borhyaenids Australohyaena antiquua, Arctodictis sinclairi, and Borhyaena macrodonta (Marshall, 1978; Forasiepi et al., 2015); and the proborhyaenid Proborhyaena gigantea (R.K.E., personal obs.). Most of the enamel is missing from p3 in SGOPV 3490, with only a small patch preserved near the apex of the left p3. It is not clear whether this paucity of enamel represents a normal feature of Eomakhaira or is an artifact of preservation, because in most other sparassodonts (e.g., Pharsophorus, borhyaenids) the enamel of p3 extends inferiorly to the same level as in the other postcanine teeth, although in Proborhyaena (AMNH 29576, MACN-A 52-382) and Callistoe (PVL 4187), the enamel of p3 is restricted to the apex of this tooth.

FIG 12.

Oblique lateral CT image slice of SGOPV 3490 along the postcanine tooth row, showing position of lower molars and depth of horizontal ramus. Note how worn surfaces of P3 and m1 closely match one another. Scale = 30 mm.

The roots of the lower postcanines of many borhyaenoids (borhyaenids, proborhyaenids, and closely related taxa) are often robust and “bulbous.” The degree to which this condition is expressed is variable, ranging from taxa with bulbous roots only on p3 (e.g., Borhyaena) to taxa in which the roots of all premolars are bulbous, but the roots of the molars are not (e.g., Thylacosmilus) to those having bulbous roots on all lower premolars and some molars (e.g., Australohyaena antiquua and most proborhyaenids). The roots of p3 in Eomakhaira are extremely robust (fig. 12) and nearly in contact, resulting in little interradicular space. Whether this condition qualifies as bulbous is uncertain, as previous studies have typically defined roots as bulbous when they are wider than the crown in occlusal view (see Forasiepi, 2009), but the crown of p3 is incompletely preserved in SGOPV 3490. Nevertheless, the morphology of the roots of p3 in SGOPV 3490 closely resembles that of sparassodonts considered to have bulbous roots, such as Arctodictis sinclairi, in which the p2–m2 roots are so swollen that the interradicular space is nearly eliminated (Forasiepi, 2009: fig. 19). The crowns of right p2 and m1 are preserved in SGOPV 3490, but their roots are not wider than their crowns, suggesting that these roots are not bulbous if prior definitions are strictly applied. The roots of right p2 are comparatively more robust than the roots of the molars (fig. 4) but still much less bulbous than in A. sinclairi. SGOPV 3490 resembles Thylacosmilus in that the roots of the lower premolars are more robust than those of the molars, though in contrast to Eomakhaira the premolar roots of Thylacosmilus are bulbous (Forasiepi, 2009).

Molars: Left M3–4 (fig. 13) and the partial crown of right M3 are the best preserved of the heavily worn upper molars. The left M3 appears to have been displaced posterolingually relative to the left M4. In most sparassodonts, the distal tip of the M3 postmetacrista contacts the anterior end of the M4 preparacrista, whereas in SGOPV 3490, the metastylar corner of M3 is located labial to the end of the M4 preparacrista. Nevertheless, the length of the M3 can be roughly estimated based on preserved parts of this tooth. The M3 is so heavily worn that it is essentially pyramidal in shape, bearing only a single, poorly distinguished cusp. This cusp is located near the labial edge of the tooth, suggesting that the stylar shelf was extremely narrow. Based on comparisons with other sparassodonts, the main cusp represents either the metacone (which is typically the tallest upper molar cusp in sparassodonts) or the remnants of a single, completely connate, merged paracone and metacone. In most sparassodonts that have a very small paracone (e.g., Borhyaena, Arctodictis, Australohyaena, Prothylacynus, and the proborhyaenid from the Tremembé Formation), this cusp is typically half (or less) the height of the metacone. However, in Callistoe, Arminiheringia, and Patagosmilus, the paracone is very tall (despite its small base), often nearly as high as the metacone (or only slightly lower). The condition in Proborhyaena is ambiguous; the relative heights of the paracone and metacone in the molars of AMNH 29576 are obscured by wear and damage.

Evidence of an anterolabial cingulum, paracone, or stylar cusps on M3 of SGOPV 3490 is lacking. Whether these structures were once present but obliterated by heavy wear cannot be determined. The ectoflexus on M3 is >10% the labiolingual width of the tooth (fig. 13B), qualifying it as “deep” sensu Davis (2007) and Williamson et al. (2012). Among short-snouted borhyaenoids, a deep ectoflexus on M3 occurs in Prothylacynus patagonicus, Proborhyaena gigantea, and Callistoe vincei (table S7). By contrast, upper molar ectoflexi are shallow or absent in Pharsophorus tenax and all borhyaenids (Borhyaena spp., Arctodictis spp., Acrocyon riggsi, Australohyaena antiquua). Although the M3 ectoflexus of Patagosmilus goini is classified here as “shallow,” it is deeper than in borhyaenids and Pharsophorus tenax (close to the threshold between “deep” and “shallow”; table S7) and contrasts with the condition in Thylacosmilus atrox and an indeterminate Colhuehuapian thylacosmiline (Goin et al., 2007), in which the ectoflexus is extremely shallow or absent. CT images show that the M3 roots are extremely splayed in SGOPV 3490. A similar condition occurs in Proborhyaena, Paraborhyaena, and MLP 79-XII-1-1. Other sparassodonts may also exhibit this condition, but this is difficult to assess without isolated molars or CT imaging data. Goin et al. (2007) considered anterolabially-posterolingually narrow molar roots to characterize Thylacosmilinae. This condition cannot be scored for Eomakhaira due to distortion of these roots in SGOPV 3490.

FIG. 13.

Left M3–4 of SGOPV 3490 in A, labial, B, occlusal, and C, lingual views (C rotated upside down for easier comparison with A, B). Anterior to left in all CT segmentation images. Scale = 3 mm.

The M3 protocone of SGOPV 3490 appears to have been extremely small. This inference is based on: (1) the position of the lingual root M3, which is mostly superior to the trigon rather than directly above it, leaving little space for a talon (unlike the condition in sparassodonts with a larger protocone); and (2) the presence of a small but poorly preserved protocone on right M3. On the left M3, the protocone is entirely missing and the lingual face is dominated by a nearly vertical surface (fig. 13B, C). It is unclear whether this feature formed in vivo or resulted from postmortem damage. The flat lingual face of the left M3 does not appear to be due to carnassial rotation, as carnassial rotation in other sparassodonts produces wear facets that parallel the preparacrista and postmetacrista (see below), whereas in SGOPV 3490, this feature is oblique to these crests. Significantly, this feature does not occur on the left M3, suggesting that its presence on the right tooth is a preservational artifact.

A small, freshly broken area of enamel on the posterolabial face of the left M3 of SGOPV 3490 shows that the enamel is remarkably thin (∼0.06 mm). The enamel seems to be of similar thickness on the lower teeth, but this is uncertain given the lack of other clean breaks and sufficient density contrast and resolution for distinguishing thin enamel from dentine in CT scans. In sparassodonts, extremely thin molar enamel has been reported in a proborhyaenid (0.17 mm in MLP 79-XII-19-1, cf. Proborhyaena; Koenigswald and Goin, 2000) and large hathliacynids (0.07–0.10 mm in Acyon and Cladosictis; Koenigswald and Goin, 2000; Engelman et al., 2015). Enamel is generally thicker in borhyaenids and basal borhyaenoids (∼0.23 mm, Prothylacynus patagonicus; ∼0.34 mm, Arctodictis sinclari; Koenigswald and Goin, 2000). Postcanine enamel thickness is unknown in thylacosmilines, but their canine enamel has been reported to be extremely thin (Turnbull, 1978; Koenigswald and Goin, 2000). Although the enamel thickness in Eomakhaira is more similar to that in other proborhyaenids than in other borhyaenoids, allometry may play a role, given that Eomakhaira is much smaller than both Prothylacynus and Arctodictis.

The best-preserved upper molar in SGOPV 3490, left M4 (figs. 13, S6), is very short anteroposteriorly. In this respect, Eomakhaira more closely resembles Patagosmilus (where the M4 is also very narrow) than other short-snouted borhyaenoids (e.g., Prothylacynus, Pharsophorus, Arminiheringia, Borhyaena, Arctodictis, and Thylacosmilus) in which the M4 is more robust and less anteroposteriorly narrow (table S8). The preparacrista is parallel to the axis of the greatest width of the tooth. The M4 crown is oriented obliquely (anterolabially-posterolingually) to the rest of the toothrow. An oblique M4 occurs in many sparassodonts (Cladosictis patagonica, Acyon myctoderos, Prothylacynus patagonicus, Callistoe vincei, Arminiheringia auceta, Arctodictis sinclairi, Arctodictis munizi, Thylacosmilus atrox) but can be variable (e.g., M4 is oriented obliquely in some but not all individuals of Arctodictis sinclairi and Acyon mycteros). The M4 of SGOPV 3490 is almost as wide or wider labiolingually than M3, even accounting for damage to the latter tooth, a pattern otherwise seen only in Pharsophorus tenax, Arctodictis sinclairi, Patagosmilus goini, and possibly Borhyaena macrodonta among short-snouted borhyaenoids.

The simple M4 crown of SGOPV 3490 consists of two poorly distinguished trigon cusps, a paracone and stylar cusp B, in addition to an extremely small protocone. There is no metacone. Both the paracone and stylar cusp B are nearly subsumed within the extremely well-developed preparacrista, though this may be exaggerated by wear. There is no anterior cingulum (preparacingulum) on M4. A labial cingulum, which occurs in some sparassodonts with simplified M4s (e.g., Prothylacynus patagonicus, MACN-A 707), is also absent in Eomakhaira. No vestigial postparacrista occurs posterior to the paracone, the presence of which has been described for Patagosmilus (Forasiepi and Carlini, 2010) and Arctodictis (Forasiepi, 2009).

FIG. 14.

Cross-sectional CT image of left M4 of SGOPV 3490 in A, oblique occlusal and B, posterior views. Shows presence of three roots. Scale bars = 5 mm.

The M4 protocone of Eomakhaira is tiny, barely a swelling of enamel on the lingual side of the paracone. This differs from the condition in Callistoe and Patagosmilus, in which the vestigial protocone is larger and more distinct from the trigon (Babot et al., 2002; Forasiepi and Carlini, 2010), as well as from Borhyaena macrodonta (MACN-A 32–390) and Pharsophorus tenax (AC 3192), in which the vestigial protocone retains a small basin. The M4 of Proborhyaena is unknown, and the M4 of Paraborhyaena could not be examined firsthand.

CT images of the left M4 of SGOPV 3490 show that this tooth has three roots despite its highly simplified morphology (fig. 14). The apices of the roots are located labially, lingually, and posteriorly, with the lingual and posterior roots merging basally, resulting in only two roots at the level of the crown. By contrast, most sparassodonts with a highly simplified (“linear”) M4 are considered to have only two roots (but see below), including nearly all short-snouted borhyaenoids for which the M4 is known (i.e., Acrocyon riggsi, Arctodictis spp., Borhyaena spp., Callistoe vincei, Paraborhyaena boliviana, Patagosmilus goini, Pharsophorus tenax, Prothylacynus patagonicus, and Thylacosmilus atrox). Australohyaena antiquua, the only short-snouted borhyaenoid with a definitively three-rooted M4, is deeply nested within a clade otherwise characterized by double-rooted M4s (Forasiepi et al., 2015). This observation, combined with the fact that the third root in M4 of SGOPV 3490 could be identified only through CT imagery, raises the question of whether some sparassodonts currently identified as having a two-rooted M4 might instead have a three-rooted M4.

Despite the senescence of SGOPV 3490, there is no evidence of carnassial rotation like that seen in some sparassodonts (see Discussion for more details). Carnassial rotation results in a unique wear pattern of completely flat wear facets with exposed dentine on the posterolingual faces of M1–3 (extending from the metastyle to the protocone) and the anterolingual face of M4 (fig. S7). Such wear facets are not seen in SGOPV 3490. Although the posterolingual corners of the left and right M3s are poorly preserved, these wear facets are clearly absent on the anterolingual face of M4. At the same time, the crowns of right M3 and left M3 and M4 appear to have been medially canted, a feature usually considered indicative of carnassial rotation (Mellett, 1969; Bond and Pascual, 1983). This canting is clearly not an artifact of postmortem distortion, as the crowns are slanted in opposite, complimentary directions on the left and right sides. Medially canted molars occur in all other borhyaenoids and hathliacynids, including many species that show no signs of carnassial rotation, even in the oldest individuals (see discussion). This demonstrates that molar canting occurs independently of, and is thus not necessarily indicative of, carnassial rotation. Whereas the posterior upper molars of Eomakhaira are medially canted, the posterior lower molars are laterally canted. Laterally canted posterior lower molars occur in other sparassodonts, such as Australohyaena (Forasiepi et al., 2015), Arminiheringia, and Arctodictis. Labial canting of the lower molars is possibly correlated with the lingual canting of the upper molars in many sparassodonts.

The lower molars of SGOPV 3490 are not strongly imbricated (fig. 15), at least not to the degree seen in the borhyaenids Australohyaena, Acrocyon, and Arctodictis, or the thylacosmiline Thylacosmilus. The m3–4 are slightly angled relative to the long axis of the toothrow, comparable in degree of imbrication to that seen in Arminiheringia and Proborhyaena but more imbricated than in Paraborhyaena, Pharsophorus (specifically the holotype, MACN-A 52-391), and possibly Borhyaena (based on MACN-A 52-366, assigned to Borhyaena macrodonta). In lateral view, the alveolar border of the lower molars, as approximated by the bases of the crowns (in the absence of most of the alveolar bone in this region), rises posteriorly at an angle of ∼5°–6° relative to horizontal. Several factors suggest this is real rather than taphonomic. First, when the specimen is positioned such that the alveolar border is horizontal, the nasals point anterodorsally, a biologically unreasonable orientation. Second, the right P3/m1 are preserved in occlusion, and the alveolus of the right P3 is partially preserved, indicating that the right dentary has not moved relative to the cranium. Finally, a lower toothrow that rises posteriorly occurs in a few other sparassodonts, including Arminiheringia auceta, Paraborhyaena boliviana, Arctodictis sinclairi, Australohyaena antiquua, and possibly Acrocyon riggsi (Goin et al., 2007; fig. 8A). In those taxa, the alveolar border of the lower molars is angled at about 8° relative to horizontal, similar to the inferred angle in SGOPV 3490.

Little can be said about m1–2 of SGOPV 3490. As mentioned above, the occlusal morphology of right m1–2 has been obliterated by wear, whereas on the left side, m1 is missing its crown, and no trace of m2 is preserved, possibly due to greater distortion of the left side of the skull. The posterior lobes of the crowns of m1–2 are not lower than the anterior lobes, a condition that occurs to a variable degree in all borhyaenids, including Prothylacynus, Plesiofelis, Pharsophorus (including P. lacerans and P. tenax but not MPEF-PV 4170, the specimen from La Cantera assigned to Pharsophorus), Proborhyaena, and Thylacosmilus. In this respect, SGOPV 3490 resembles Arminiheringia. A small posterolabial cingulid occurs on the labial side of m1 in SGOPV 3490 (fig. 4). The presence or absence of a posterolabial cingulid cannot be evaluated in m2 or m3, but it is absent on m4. The presence of a labial postcingulid on m1 is not unexpected given its distribution among sparassodonts. Posterolabial cingulids are absent in the basal borhyaenoids Lycopsis, Pseudothylacynus, and Prothylacynus but are present in cf. Nemolestes (AMNH 29433; Forasiepi et al., 2015), Plesiofelis, Pharsophorus (both P. lacerans and P. tenax), all borhyaenids (Acrocyon spp., Australohyaena antiquua, Arctodictis spp., and Borhyaena spp.), and the proborhyaenid Callistoe (Forasiepi et al., 2015). This feature could not be scored for Arminiheringia, Proborhyaena, or Paraborhyaena. The condition in thylacosmilines is not entirely clear. Photographs of the holotype of Anachlysictis gracilis in Goin (1997) appear to show a posterolabial cingulid, whereas figures of Thylacosmilus (Riggs, 1934; Marshall, 1976a) show a structure that could be a posterolabial cingulid. Most phylogenies of sparassodonts (Forasiepi, 2009; Engelman and Croft, 2014; Forasiepi et al., 2015; Suarez et al., 2016) imply at least six independent losses of the posterolabial cingulid (in Hondadelphys, Stylocynus, Lycopsis, Prothylacynus, and at least twice among hathliacynids, but see Phylogenetic Analysis below), or repeated loss and reacquisition of the posterolabial cingulid within Sparassodonta.

FIG. 15.

Lower right dentition of SGOPV 3490 in A, labial, B, occlusal, and C, lingual views. Colors of elements in this CT segmentation are the same as in figure 3. Anterior to right in A, B and to left in C. Scale = 30 mm.

The right m4 is the best-preserved lower molar in SGOPV 3940 and the only tooth in which crown morphology has not been largely obliterated by wear. Nevertheless, its paracristid still is highly worn, comparable to the degree of wear in the holotype of Angelocabrerus daptes (MMP 967M; Simpson, 1970) and a specimen assigned to Pharsophorus cf. P. lacerans (MPEF-PV 4190; Goin et al., 2010). As in all borhyaenoids, the m4 of SGOPV 3490 is characterized by two main cusps, a tall protoconid and a slightly shorter paraconid. The m4 paraconid of SGOPV 3940 lacks an anteriorly projecting ventral keel, unlike most sparassodonts but as in proborhyaenids and possibly Thylacosmilus. The protoconid is tall relative to the anteroposterior length of the tooth (i.e., the height is greater than 90% the length of the tooth; see Muizon et al., 2018) and is wider at its midpoint that at its base, as in other sparassodonts. The m4 of SGOPV 3490 closely resembles that of Proborhyaena and Paraborhyaena in having a posteriorly salient protoconid at the posterior end of the tooth and a barely discernable talonid. The latter feature contrasts with m1–3, each of which shows evidence of a talonid that is very small and worn but nevertheless slightly larger.

The metaconid is clearly absent on the m4 of SGOPV 3490 (fig. 15). Assessing whether a metaconid was present or absent on m2–3 is more difficult due to the worn and highly fragmented preservation of these teeth, but this cusp appears to be absent on at least m3. In borhyaenoids with a metaconid (e.g., borhyaenids, Pharsophorus), this cusp often occurs as a small, low protuberance at the posterolingual corner of the tooth (see Forasiepi et al., 2015: figs. 10, 11). In SGOPV 3490, on the other hand, the posterolingual surface of m3–4 is smooth, and the base of the protoconid extends to the lingual margin of the tooth; there is no evidence that a distinct metaconid was present. In fact, the posterior margin of m3 is very similar to the holotype of Arminiheringia auceta (MACN-A 10970), consisting of a cuspless ridge that is oriented dorsolingually-ventrolabially.

The anterior root of the posterior lower molars (primarily m3–4) in SGOPV 3490 is much larger and more robust than the posterior one. This condition is also seen in several other proborhyaenids and borhyaenids, including Borhyaena macrodonta, Borhyaena tuberata, Arctodictis sinclairi, Arctodictis munizi, Acrocyon riggsi, Acrocyon sectorius, MLP 88-V-10-4 (the proborhyaenid from Antofagasta de la Sierra), Arminiheringia auceta, Paraborhyaena boliviana, Proborhyaena gigantea, and Thylacosmilus atrox (table S9). However, this condition is not present in the borhyaenid Australohyaena antiquua, the proborhyaenid Callistoe vincei, and all species of non-proborhyaenid, non-borhyaenid sparassodonts in which the state of the roots of the lower molars could be determined (e.g., Hondadelphys, hathliacynids, Pharsophorus spp., Prothylacynus patagonicus; see table S9). Disparity in size between the anterior and posterior roots varies among taxa. In Arctodictis sinclairi and Arminiheringia auceta, only the roots of m3–4 are unequal in size, whereas in Arctodictis munizi, Proborhyaena gigantea, Paraborhyaena boliviana, and Thylacosmilus atrox, it is the roots of m2–4 that are unequal. In SGOPV 3490, the anterior roots of m2–4 are larger than the posterior ones, but the disparity is much less in m2 than in m3–4. Molar roots differ in size among other groups of carnivorous mammals. The roots of m2–3 are similar in length and robustness in the extant bone-cracking Sarcophilus harrisii, while the anterior root of m4 is slightly more robust than, but the same length as, the posterior one (Fiani, 2015). This size disparity is also present in the lower carnassial (m1) of some carnivorans that are not specialized bone-crackers, including some species of barbourofelids (Tseng et al., 2010), felids, and Cryptoprocta (R.K.E., personal obs.).

Paleobiology of SGOPV 3490

Body Mass: Compared to other borhyaenoids, Eomakhaira molossus is notable for its small size. Regression equations of the lower dentition from Myers (2001) and Gordon (2003) yield body mass estimates of 9.5–10 kg for Eomakhaira molossus (table S10), comparable to a male Tasmanian devil (Sarcophilus harrisii; Rose et al., 2016). The holotype specimen, SGOPV 3490, is comparable in size to a skull of Sarcophilus, though SGOPV 3490 is deeper and narrower. This unusual shape is unlikely to be attributable to postmortem compression. While the skull has been subjected to shear deformation, there are no signs of significant dorsoventral or mediolateral crushing.

Eomakhaira is by far the smallest known Paleogene proborhyaenid (fig. 19, table S11), its estimated body mass being only ∼40% of that of the next smallest, Callistoe vincei (∼23 kg; Argot and Babot, 2011). Such a small proborhyaenid is quite unexpected during the Oligocene, given that non-thylacosmiline proborhyaenids reached their largest size during this interval. The two previously described Oligocene proborhyaenids, Proborhyaena and Paraborhyaena, both likely exceeded 50 kg in body mass (Zimicz, 2012; Prevosti et al., 2013; Croft et al., 2018). The small size of Eomakhaira indicates that proborhyaenids exhibited significantly more ecomorphological diversity than previously recognized during the Oligocene. This conclusion is corroborated by an unnamed proborhyaenid from the late Oligocene (Deseadan SALMA) locality of Taubaté (Couto-Ribeiro, 2010) that is comparable in size to Arminiheringia auceta (table S11), making it much larger than Eomakhaira but still smaller than Paraborhyaena and Proborhyaena (probably about 30–40 kg based on comparison with A. auceta; Prevosti et al., 2013; Croft et al., 2018). There are two possible explanations for the high ecomorphological diversity of proborhyaenids during the Oligocene. First, Oligocene proborhyaenids could represent holdovers from a much older unsampled Eocene radiation, of which only a few lineages (such as those that gave rise to Proborhyaena + Paraborhyaena and Thylacosmilinae) survived into the Oligocene. Alternatively, the high Oligocene diversity of Proborhyaenidae might reflect an Oligocene radiation related to faunal turnover at the Eocene-Oligocene transition, of which gigantism and small size/incipient machairodonty were but two evolutionary experiments. South American metatherians underwent a large faunal turnover during the Oligocene, likely in response to climatic change. This turnover (termed the Bisagra Patagónica) has been associated with the Eocene-Oligocene transition (EOT) by previous authors and has been considered analogous to the Grande Coupure in Europe and the Mongollian Remodelling in Asia (Goin et al., 2010; 2016; Abello et al., 2020). However, the peak of the turnover does not correlate with the EOT (and the faunal events associated with it on other continents) but instead occurs nearly 4 million years later, during the “middle” Oligocene (∼30.6–30.7 Ma; “Canteran”; see Goin et al., 2010; Dunn et al., 2013). Regardless of its exact correlation with the EOT, this Oligocene turnover resulted in the extinction of many formerly dominant metatherian groups and subsequent radiation of the survivors (Goin et al., 2010; 2016; Abello et al., 2020). Among sparassodonts, these changes included the appearance of hathliacynids and borhyaenids in the late Oligocene (Petter and Hoffstetter, 1983; Forasiepi et al., 2015). The extinction of non-thylacosmiline proborhyaenids at the end of the Oligocene has been considered an example of the long-term decline of a previously dominant group stemming from climatic changes during the Oligocene (Goin et al., 2016). However, given that gigantism and machairodonty do not occur in middle Eocene proborhyaenids such as Callistoe or Arminiheringia, proborhyaenids may have radiated in response to the environmental changes and faunal turnover during the Oligocene in a manner similar to hathliacynids and borhyaenids. The fossil record of proborhyaenids is too incomplete at present to favor either hypothesis, though they both represent avenues for future research.

FIG. 19.

Size comparison of representative Paleogene proborhyaenids. From largest to smallest, Proborhyaena gigantea (in blue), the largest known proborhyaenid (scaled after AMNH 29576, the largest specimen of this taxon); Callistoe vincei (in green), the smallest named proborhyaenid prior to this study (scaled after the holotype specimen, PVL 4187); Eomakhaira molossus (in red), scaled after SGOPV 3490. Homo sapiens (170 cm tall) to right for comparison. Body mass for Proborhyaena from Prevosti et al. (2013) and Croft et al. (2018), Callistoe from Argot and Babot (2011), and Eomakhaira from the present study. Silhouettes for Proborhyaena (CC-BY-SA 3.0), duplicated for Callistoe and Eomakhaira, and Homo (CC0 1.0) by Zimices, and NASA, respectively, from PhyloPic. Scale bar = 1 m.

The small size of Eomakhaira raises the possibility that an isolated m4 from the late Eocene (Mustersan) of Antofagasta de la Sierra in northwestern Argentina (MLP 88-V-10-4) belongs to the same taxon or a closely related form. Originally assigned to Arminiheringia by Goin et al. (1998), later authors (Babot et al., 2002; Babot, 2005; Powell et al., 2011) have referred this specimen to Callistoe based on its smaller size and reduced talonid. Although MLP 88-V-10-4 is larger than the corresponding tooth of the holotype of Eomakhaira (m4 length is 14.0 mm in MLP 88-V-10-4 versus 12.0 mm in SGOPV 3490), it is smaller than the corresponding tooth in the holotype of Callistoe (length = 17 mm; Babot et al., 2002). MLP 88-V-10-4 also more closely resembles SGOPV 3490 in the more posterior position of its protoconid and near absence of a talonid; in Callistoe, the protoconid is positioned more anteriorly and is less salient posteriorly, and the talonid is larger. Furthermore, MLP 88-V-10-4 lacks an anterolabial cingulid (Goin et al., 1998: fig. 7A), similar to Proborhyaena, Paraborhyaena, and Thylacosmilus (but apparently not the thylacosmiline Anachlysictis, see Goin, 1997: fig. 11.6B). This suggests that MLP 88-V-10-4 cannot be assigned to either of the currently recognized genera of Eocene proborhyaenids, Callistoe or Arminiheringia, in which an anterolabial cingulid is present on m4 (Babot et al., 2002). While the size and morphological resemblance support a potential close relationship with Eomakhaira, this is not definitive, as the state of the anterolabial cingulid cannot be determined in SGOPV 3490.

The small size of Eomakhaira relative to other proborhyaenids suggests that thylacosmilines were ancestrally characterized by small body size compared to other members of this group. Eomakhaira is small (9.5–10 kg) not only relative to Paleogene proborhyaenids but also to many late Cenozoic thylacosmilines (table S11). The proborhyaenid taxa stemward of the base of Thylacosmilinae are all relatively large (>20 kg), with the smallest of these species, Callistoe, being two and a half times larger than Eomakhaira. The only members of Proborhyaenidae (including thylacosmilines) comparable to Eomakhaira in size (table S11) are other geologically old thylacosmilines such as the Colhuehuapian thylacosmiline from Gran Barranca and the possible plesiomorphic thylacosmiline from La Venta (the latter being substantially smaller than Eomakhaira). The slightly larger Patagosmilus goini is estimated to have been less than 20 kg (Prevosti et al., 2013). If MLP 88-V-10-4 from the late Eocene of Antofagasta de la Sierra also pertains to Eomakhaira or a closely related form, it would further support the idea that thylacosmilines were ancestrally characterized by small body size, as MLP 88-V-10-4 is slightly larger than SGOPV 3490, potentially implying a decrease in size across the Eocene-Oligocene boundary. The large body size and highly specialized morphology of non-thylacosmiline proborhyaenids has previously been used to argue against a close relationship between thylacosmilines and proborhyaenids sensu stricto (Simpson, 1948; Bond and Pascual, 1983), as large, morphologically specialized carnivorous mammal clades generally exist for short times-pans (<10 million years) rather than continue to diversify for millions of years (Van Valkenburgh, 1999; Van Valkenburgh et al., 2004). Ancestrally small body size in thylacosmilines may have circumvented this pattern and allowed for a later, secondary acquisition of large body size in this clade later in the Cenozoic.

Dietary Habits: Several features suggest that borhyaenids and non-thylacosmiline proborhyaenids used p3, the largest and most robust premolar in all sparassodonts, to crack bones. These include an interlocking or fused mandibular symphysis, deep dentary, bulbous premolars (especially P/p3) with long roots, enamel microfractures, and high estimated bite force (Blanco et al., 2011; Ercoli et al., 2014; Forasiepi et al., 2015; Echarri et al., 2017). Contrary to a possible first impression of this robust-skulled new taxon, Eomakhaira does not seem to have been specialized for bone cracking (sensu Werdelin, 1989). Previous authors have noted that mandible depth in bone-cracking carnivorous mammals is often that same below the primary carnassial and the bone-cracking premolar, resulting in the ventral border of the horizontal ramus of the dentary to appear straight in lateral view (Palmqvist et al., 2011; Forasiepi et al., 2015). More specifically, Palmqvist et al. (2011) considered a uniformly deep mandible with a straight ventral border between the bone-cracking premolar and primary carnassial to distinguish obligate scavenging hyaenids (Hyaena hyaena and Parahyaena brunnea) from facultative scavengers that also hunt for prey (i.e., Crocuta crocuta, wherein the dentary is deeper beneath the carnassial than the bone-cracking premolar). This observation may not be broadly applicable, however, given that it is only based on extant hyaenids; extinct hyaenid species may have differed in their dietary habits despite similar mandible morphology (DeSantis et al., 2017). Moreover, since Hyaena and Parahyaena are considered sister taxa among extant hyaenids (e.g., Koepfli et al., 2006), it is not clear whether their similar mandible shape reflects functional morphology or shared ancestry. In extant metatherians, a uniformly deep dentary with a straight ventral border between the primary crushing tooth and carnassial is present not only in bone-cracking Sarcophilus but also in non-bone-cracking Dasyurus. Thus, this feature cannot be used to securely distinguish obligate bone-cracking scavengers from facultative bone-crackers.

In sparassodonts, a deep mandible with a straight ventral border and roughly uniform depth between p3 and m4 occurs in Australohyaena, Proborhyaena, Paraborhyaena, and Arctodictis, and these taxa have all been considered to be bone-crackers (Blanco et al., 2011; Zimicz, 2012; Ercoli et al., 2014; Forasiepi et al., 2015; Echarri et al., 2017). By contrast, the dentary is much shallower under p3 than under m4 in Eomakhaira, and the ventral border of the mandible is curved in lateral view. The symphysis of Eomakhaira is also much less extensive than in presumed bone-cracking sparassodonts. In most such sparassodonts, the symphysis extends posteriorly to at least the main bone-cracking premolar, often to the p3/m1 embrasure and sometimes to the posterior root of m1 (in Arminiheringia auceta; Babot et al., 2002; Zimicz, 2012). In the extant bone-cracking dasyuromorphian Sarcophilus, the symphysis extends to the midpoint of m1, which is functionally analogous to the p3 of sparassodonts. In Eomakhaira, the symphysis ends approximately at the level of the p2/p3 embrasure. Also, the roof of the skull is not vaulted at the level of the primary bone-cracking teeth in Eomakhaira, in contrast to the typical condition in other bone-cracking mammals (Werdelin, 1989), including Australohyaena, in which the nasals are vaulted (Forasiepi et al., 2015).

Zimicz (2012; 2014) and Forasiepi et al. (2015) used several metrics modified from Van Valkenburgh (1989) to examine dietary habits in carnivorous metatherians, primarily degree of specializations for bone-cracking (durophagy) and/or a carnivorous diet (i.e., hypercarnivory, mesocarnivory, etc.). Five of these parameters (premolar shape, relative premolar length, relative premolar size, relative trigonid length, and relative grinding area) can be applied to Eomakhaira (table 4). Premolar shape of the putative bone-cracking tooth (width/length of p3) reflects the robustness of the last lower premolar; it is 0.54 in Eomakhaira, slightly below the threshold separating bone-cracking from non-bone-cracking forms (bone-crackers >0.58). Relative premolar size (width of p3/cube root of body mass in kg) is 1.32, which is substantially lower than the threshold between bone-cracking and non-bone-cracking taxa (bone-crackers >2.6). The relative premolar length (length of p3/length of m4) is 0.63, below the threshold of hypercarnivores (hypercarnivores >0.7). However, extant metatherian hypercarnivores such as Sarcophilus harrisii and Dasyurus maculatus (see Maga and Beck, 2017), as well as many fossil sparassodonts exhibiting other specializations considered related to hypercarnivory (Callistoe, Arminiheringia, Australohyaena), also fall below this threshold, suggesting that this parameter is not applicable across metatherians.

With regard to the molars, the relative trigonid length (m4 trigonid length/total m4 length) of Eomakhaira is 0.91, within the range of specialized (“catlike”) hypercarnivores and comparable to some of the most specialized carnivores within Sparassodonta, including the borhyaenoid Angelocabrerus; the borhyaenids Australohyaena, Arctodictis, and Acrocyon; the proborhyaenids Proborhyaena and Paraborhyaena; and the thylacosmiline Thylacosmilus (Zimicz, 2012; Forasiepi et al., 2015; Croft et al., 2018). Zimicz (2012) and Forasiepi et al. (2015) considered a relative trigonid length of >0.9 to indicate catlike hypercarnivory and a relative trigonid length of 0.8–0.9 to indicate hypercarnivory with bone-crushing specializations. However, only three extant bone-crushing taxa (Crocuta crocuta, Hyaena hyaena, and Parahyaena brunnea; all placental hyaenids) were included in the comparative dataset of those studies. Hyaena and Parahyaena are unusual among extant large-bodied hypercarnivores in having small but functional talonids on m1 (Ewer, 1954), whereas Crocuta more closely resembles felids in its longer trigonid and a near-vestigial talonid (and has a relative trigonid length >0.9; Van Valkenburgh, 1989). This difference may be related to the substantial amount of fruit consumed by Parahyaena and Hyaena (∼12%–20% of diet by volume) but not by Crocuta (Kruuk, 1976; Owens and Owens, 1978; Mills, 2015). Molar trigonid length is unlikely to be correlated with bone-cracking habits in hyaenids, as hyenas primarily employ premolars in this function (Werdelin, 1989). Additionally, the relative trigonid lengths of the extant non-bone-cracking hypercarnivorous canids Cuon, Lycaon, and Speothos are only 0.72–0.74 (Van Valkenburgh, 1989). The narrow range of relative trigonid lengths in living bone-cracking carnivorans (0.8–0.9) most likely reflects their low diversity (N = 4, the three hyaenids and Sarcophilus harrisii) and phylogenetic signal rather than any unique functional association; three of the four extant bone-crackers (Crocuta crocuta, Hyaena hyaena, and Parahyaena brunnea) belong to the same clade, and this range is within the variation seen in mammalian hypercarnivores more generally.

TABLE 4.

Morphometric values of the dentition used to infer dietary habits in Eomakhaira molossus. Methodology for calculating these parameters and critical values for dietary categories based on Van Valkenburgh (1989), Zimicz (2012), and Forasiepi et al. (2015).

Relative grinding area (RGA) has been assessed via two methods in sparassodonts, one using only m4 (Prevosti et al., 2013) and the other assessing the entire lower molar row (Croft et al., 2018). Although the latter method is optimal (see Croft et al., 2018), m1–3 in the holotype of Eomakhaira are too heavily worn to calculate RGA in this manner. The m4 of Eomakhaira most closely resembles that of Proborhyaena, Paraborhyaena, Arctodictis, and Thylacosmilus, all of which were characterized by Prevosti et al. (2013) as lacking a functional talonid and therefore considered to have an RGA of 0. Evaluated together, the short mandibular symphysis, shallow dentary below p3, labiolingually narrow premolars that are small in proportion to those of bone-cracking taxa, long trigonid, and very small talonid of Eomakhaira suggest it was hypercarnivorous but not clearly specialized for bone crushing.

Carnassial Rotation in Sparassodonts

Mellett (1969) described an unusual phenomenon in the hyaenodont Hyaenodon in which the upper molars become progressively more medially oriented throughout ontogeny (carnassial rotation). Mellett (1969) also reported the occurrence of carnassial rotation in the hyaenodont Hemipsalodon, the oxyaenid Patriofelis, and “an unnamed Pliocene marsupial saber-tooth” (Mellett, 1969; almost certainly referring to Thylacosmilus atrox, given that no other well-preserved thylacosmilines were known at the time). Carnassial rotation has been documented in extinct carnivorans, including the barbourofelid Barbourofelis (see Baskin, 1981) and the nimravid Hoplophoneus (see Bryant and Russell, 1995). Marshall (1978) reported carnassial rotation in several sparassodonts, most prominently Arminiheringia auceta, but also to a lesser degree in Acrocyon, Arctodictis, and Borhyaena (but see below). Bond and Pascual (1983) described carnassial rotation in a senescent specimen they assigned to Proborhyaena (MLP 79-XII-18-1). Carnassial rotation in the placental Hyaenodon and the metatherian Arminiheringia produces a very distinct form of wear in which the entire posterolingual face of the upper molars is worn flat and the dentine is exposed, even on the talon (fig. S7). This produces a sharp, flat edge with vertical wear facets that roughly parallels the main shearing blade of the tooth (typically the postmetacrista of upper molars) but relatively little apical wear. In the holotype of Arminiheringia auceta (MACN-A 10970/10972), in vivo wear had progressed to the point that the pulp cavities of M1–3 were exposed. Additionally, in the holotype of A. auceta, a second flat wear facet is present on the anterolingual face of M1–4, roughly parallel to the preparacrista. These facets are not present in Hyaenodon (M. Borths, personal commun.).

Aside from Arminiheringia, the only other sparassodonts for which true carnassial rotation appears to be present are the proborhyaenid Callistoe vincei and thylacosmiline Thylacosmilus atrox. The holotype of Callistoe vincei (PVL 4187) exhibits the same vertically flat wear facets seen in Arminiheringia on the entire posterolingual faces of M2–3 and anterolingual faces of M2–4 (M1 not being preserved in this specimen), including the talons. In Thylacosmilus, FMNH P14531 exhibits vertically flat wear facets on the posterolingual face of M3 and anterolingual faces of M4 (as determined from a cast of this specimen). Goin and Pascual (1987) found no evidence of ontogenetic carnassial rotation in Thylacosmilus, perhaps because they examined ontogenetically younger specimens (with less-worn teeth) than FMNH P14531, which pertains to an older individual with a highly worn dentition. Contra Marshall (1978), we have not observed carnassial rotation in any borhyaenid. The dentitions of borhyaenids are not characterized by flat wear facets across the entire lingual faces of the teeth; rather, wear facets are predominantly apical and follow the major cusps and crests of the crowns, as in most other faunivorous metatherians (e.g., Didelphis; see Crompton and Hiiemae, 1970). Bond and Pascual (1983) reported carnassial rotation in cf. Proborhyaena (MLP 79-XII-18-1) based on lingual canting of the upper molars and labial canting of the lower molars. However, as discussed below, neither of these features is necessarily indicative of carnassial rotation. The molars of MLP 79-XII-18-1 are too damaged to determine whether flat wear facets similar to those seen in Arminiheringia and Thylacosmilus were present. Carnassial rotation does not characterize all proborhyaenids despite its presence in Callistoe, Arminiheringia, and Thylacosmilus. Upper molars of Proborhyaena gigantea (AMNH 29576) clearly do not exhibit carnassial rotation. Wear facets on the upper molars of AMNH 29576 are restricted to their occlusal faces, as in most other sparassodonts.

Although only a handful of sparassodonts exhibit true carnassial rotation, many sparassodonts do exhibit upper molars that are medially canted. This condition can be identified by measuring the angle between the bases of the crowns of the posterior upper molars (M3–4) and the horizontal plane in posterior (distal) view; we consider an angle >35° indicative of canted molars. In general, the posteriormost upper molars (M3–4) show the highest degree of canting, whereas more anterior molars (e.g., M1) are less canted and often comparable in orientation to the canines and premolars. In taxa with canted upper molars, the paracone and metacone are typically canted medially, and the occlusal faces of the upper molars are partially or completely obscured when the palate is in ventral view, sometimes to the point that the protocone is hidden by the trigon (figure S5 illustrates this condition in borhyaenids). In effect, this means that the upper molars are not oriented in occlusal view when the palate is viewed ventrally, something that can readily be seen when sparassodont skulls are figured in ventral view (Sinclair, 1906: pls. 42, 44, 48, 55, 59, 60; Babot et al., 2002: figs. 5C, 6B; Forasiepi, 2009: fig. 20; Forasiepi et al., 2015: figs. 2, 6A). In addition to SGOPV 3490, molars canted at angles >35° are present in the hathliacynids Acyon myctoderos, Cladosictis patagonica, and Sipalocyon gracilis, and in the borhyaenoids Acrocyon riggsi, Arctodictis sinclairi, Prothylacynus patagonicus, Pharsophorus tenax, and Arminiheringia sp. (table 5; fig. S8). By contrast, the angle between the palate and the crowns of M3–4 is 10°–25° in Allqokirus australis, Patene simpsoni, Patene coluapiensis, Hondadelphys fieldsi, and UF 27881 (as indicated by the alveoli of M2–3). Both states contrast with the condition in metatherians such as Didelphis, where the crown bases lie nearly level with the palate (angle <10°). The upper molars are also medially canted in Australohyaena antiquua, Borhyaena spp., Callistoe vincei, cf. Proborhyaena, Paraborhyaena boliviana, and Thylacosmilus atrox, but the precise angle cannot be determined from available photographs and published illustrations (which typically do not show the palate in posterior view). All undoubted members of Hathliacynidae + Borhyaenoidea that could be observed directly exhibit strongly medially canted upper molars.

Although the upper molars of many sparassodonts are medially canted, they do not appear to have undergone carnassial rotation as seen in Hyaenodon and Arminiheringia. Few sparassodonts exhibit the distinctive wear pattern observed in Arminiheringia, even in older individuals, suggesting that molar orientation changed little after eruption in those taxa. This interpretation is supported by MACN-A 5931 (Prothylacynus patagonicus) and MLP 82-V-1-1 (an undescribed specimen referred to Arminiheringia sp.), both subadults with M/m4 still erupting and little to no wear on M3 (Forasiepi and Sánchez-Villagra, 2014). In these individuals, M3 is already medially canted at an angle comparable to that seen in adult specimens, and M4 is erupting in a canted orientation (fig. S8B). Additionally, MLP 82-V-1-1 bears flat wear facets on the posterolingual faces of M1–2 similar those of the (adult) holotype of Arminiheringia (Forasiepi and Sánchez-Villagra, 2014: fig. 2B), indicating that carnassial rotation can be identified early in ontogeny in sparassodonts for which this condition is present. Thus, the upper molars of most sparassodonts appear to have erupted in a canted position rather than having rotated into that position later in life.

The morpho-functional significance of medially canted upper molars in sparassodonts is unclear, given that it appears to be uncorrelated to carnassial rotation. Mellett (1969) suggested that carnassial rotation, as occurs in Hyaenodon, maintained precise occlusion between the carnassials despite wear and prolonged the functional lifespan of the dentition in this taxon. According to Mellett (1969), a carnivorous mammal like Hyaenodon, with multiple shearing teeth, anisognathus lower jaws, and fused mandibular symphysis, cannot compensate for tooth wear via lateral motion of the lower jaw (as occurs in forms with unfused symphyses, such as most extant carnivorans), and carnassial rotation provided an analogous functional solution. However, medially canted upper molars occur in many sparassodonts with a ligamentous symphysis (e.g., Acyon, Cladosictis, Sipalocyon), which provides some symphyseal flexibility. The same is true for Barbourofelis and nimravid carnivorans. Although Baskin (1981) suggested that the long vertical symphysis of Barbourofelis would have been functionally equivalent to a fused symphysis, Bryant and Russell (1995) suggested that this taxon was capable of lateral lower jaw mobility. Furthermore, as noted, the upper molars of immature specimens of Arminiheringia sp. and Prothylacynus patagonicus are medially canted, M4 erupts in a canted orientation, and the angle between the base of the molar and the palate in these specimens is comparable to that of adult sparassodonts (including other specimens of the same taxa). This indicates that medial canting of the upper molars was not attained gradually over an animal's lifespan, as might be expected if canting maintained precise occlusion during wear, since M3–4 are already canted while they erupt. Finally, the lower molars of many sparassodonts are laterally canted (e.g., Australohyaena antiquua; see Forasiepi et al., 2015), a condition also present to a lesser degree in Dasyurus (see Macrini, 2005a). The lower molars of Hyaenodon are not canted, suggesting that the functional explanation for carnassial rotation in Hyaenodon by Mellett (1969) may not be applicable to sparassodonts (except possibly to Arminiheringia, Callistoe, and Thylacosmilus). Bryant and Russell (1995) suggested that canting in Dinictis and carnassial rotation in Hoplophoneus (medial canting and carnassial rotation do not cooccur in these taxa) counter heavy wear stemming from the dental morphology and more posterior location of the carnassials of these taxa compared to other carnivorans. Such reasoning would not apply to most metatherian carnivores (except Thylacosmilus and possibly Patagosmilus; Goin, 1997; Forasiepi and Carlini, 2010), whose primary carnassials (m4) are located at the anteroposterior midpoint of the mandible (Werdelin, 1987), as in carnivorans.

TABLE 5.

Angle of medial canting of posterior upper molars in sparassodonts. Inward canting angle measured as angle between bases of M3–4 and palate in posterior view.

The mammaliaform Morganucodon exhibits medially canted posterior upper molars and uncanted anterior molars (Jäger et al., 2019), similar to what is described here for sparassodonts. Morganucodon also resembles many sparassodonts in having prominent pits on the maxilla that receive the main cusp of the lower molars and posterior upper molars positioned lingual to the lateral edge of the maxilla (i.e., “maxillary cheeks”), which suggest analogous jaw mechanics. These features have been argued to minimize “roll” (sensu Grossnickle, 2017) during mastication in Morganucodon, allowing precise occlusion despite mainly orthal jaw movements. However, canting of the upper molars in Morganucodon is interpreted to have increased tooth wear rates (Jäger et al., 2019), which runs counter to the view that upper molar canting prolongs the functional lifespan of the dentition (Mellett, 1969; Bryant and Russell, 1995).

Alternatively, canting of the posterior molars in sparassodonts may be related to their extremely tall lower molar protoconids compared to those of other carnivorous metatherians (Muizon et al., 2018). Canting may be a way to accommodate these tall teeth when the mouth is closed. The embrasure pits on the palate of many sparassodonts (e.g., Engelman and Croft, 2014) have been suggested to represent a similar adaptation to accommodate the extremely tall protoconids (Forasiepi, 2009; Muizon et al., 2018).

Proborhyaenidae and the Origin of Thylacosmilines

Implications of Eomakhaira molossus for Understanding the Evolution of Thylacosmilinae: Ever since the first well-preserved specimens of thylacosmilines were described (Riggs, 1933, 1934), the manner in which these animals acquired their distinctive, highly specialized machairodont morphology has been debated, as has their relationship to other sparassodonts. Based on various similarities, Scott (1937) tentatively linked Thylacosmilus and the Eocene proborhyaenid Arminiheringia, but this idea was questioned based on substantial morphological and temporal differences between Thylacosmilus and non-thylacosmiline proborhyaenids (Simpson, 1948; Marshall, 1976a; Bond and Pascual, 1983). Early workers were hampered in their attempts to link thylacosmilines to other sparassodonts because the only well-known member of the group was Thylacosmilus atrox, its geologically youngest and most autapomorphic member, which shared few features with other taxa known at the time. Only much later did earlier-diverging forms with less extreme machairodont specializations come to light. Goin (1997) described two taxa from the middle Miocene locality of La Venta, Colombia, (Laventan SALMA) exhibiting less machairodont specializations than the Mio-Pliocene Thylacosmilus: the thylacosmiline Anachlysictis gracilis and a second taxon, questionably assigned to the group (IGM 251108). More recently, even older thylacosmiline remains have been described from the early Miocene (Colhuehuapian SALMA; Goin et al., 2007) and early middle Miocene (Colloncuran SALMA; Forasiepi and Carlini, 2010) of Patagonia. Paradoxically, these older Patagonian taxa appear to be more closely related to Thylacosmilus than to the taxa from La Venta, even though Thylacosmilus and the La Venta taxa are closer in age; this implies an older (likely pre-Miocene) origin of the clade (Goin et al., 2007; 2016).

Our phylogenetic analysis recovers the early Oligocene Eomakhaira as the basalmost member of a clade that includes the thylacosmilines Patagosmilus and Thylacosmilus, which collectively are nested within Proborhyaenidae. The idea of a close relationship between thylacosmilines and proborhyaenids sensu stricto is not novel. Several studies (Marshall et al., 1990; Muizon, 1999; Babot et al., 2002) have recovered proborhyaenids sensu stricto and thylacosmilines as sister groups, and several others (Babot, 2005; the equal-weights analysis of Forasiepi et al., 2015; Suarez et al., 2016; Muizon et al., 2018) have recovered Thylacosmilinae within a paraphyletic Proborhyaenidae. Carneiro (2018) recovered a paraphyletic Proborhyaenidae, with Paraborhyaena as sister group to Thylacosmilinae, but also found Borhyaenidae nested within this group, with Callistoe and Arminiheringia recovered as basal to a clade composed of Borhyaenidae + (Paraborhyaena + Thylacosmilinae). Support for a close relationship between Thylacosmilinae and Proborhyaenidae sensu stricto has not been universal in previous studies, and placements for Thylacosmilinae outside of Proborhyaenidae have also been suggested (Patterson and Marshall, 1978; Bond and Pascual, 1983; Goin, 1997, 2003; Forasiepi, 2009; Engelman and Croft, 2014; the implied-weights analysis of Forasiepi et al., 2015). Recovery of Eomakhaira as an Oligocene thylacosmiline not only breaks up the long ghost lineage between thylacosmilines and non-thylacosmiline proborhyaenids but also corroborates the nesting of thylacosmilines within Proborhyaenidae, given that Eomakhaira exhibits a combination of derived features occurring in other thylacosmilines and plesiomorphic features retained in non-thylacosmiline proborhyaenids.

Eomakhaira resembles non-thylacosmiline proborhyaenids and differs from other thylacosmilines in retaining lingual sulci on the upper canines, three premolars, replacement of dP3, and absence of a genial flange. Its P1 is asymmetric, which is likely ancestral for proborhyaenids given its presence in Callistoe and Arminiheringia; the tooth is unknown in Proborhyaena and Paraborhyaena and is interpreted as absent/lost in other thylacosmilines. SGOPV 3490 lacks enamel on the labial surface of the canines, which suggests that enamel, if present in Eomakhaira, was a simple cap lost through wear as in non-thylacosmiline proborhyaenids rather than a persistant band extending to the base of the tooth as in Patagosmilus and Thylacosmilus (Turnbull, 1978; Forasiepi and Carlini, 2010; Koenigswald, 2011), Finally, Eomakhaira may have had open-rooted (hypselodont) lower canines, as in non-thylacosmiline proborhyaenids but unlike thylacosmilines. Its lower molars lack an anteriorly projecting ventral keel on the paraconid, as in non-thylacosmiline proborhyaenids (present in Anachlysictis but evidently absent in Thylacosmilus among thylacosmilines).

On the other hand, Eomakhaira has canines that lack longitudinal grooves, as other thylacosmilines (and, possibly, Lycopsis viverensis). The upper canines are narrow labiolingually compared to other sparassodonts, have a well-developed median keel, and lack a labial median sulcus. The maxilla of Eomakhaira is deep, as in other thylacosmilines (a condition acquired independently in borhyaenids), rather than shallow as in proborhyaenids. The dentary of Eomakhaira is intermediate in depth between those of other thylacosmilines and non-thylacosmiline proborhyaenids, shallower than in the Eocene proborhyaenids Callistoe and Arminiheringia (though possibly not the Oligocene Paraborhyaena and Proborhyaena) but deeper than in the thylacosmilines Anachlysictis and Thylacosmilus. The infraorbital foramen of Eomakhaira is located at the P3/M1 embrasure, a position more similar to thylacosmilines than most non-thylacosmiline proborhyaenids (except possibly Proborhyaena), in which it is located more anteriorly. The mandibular symphysis of Eomakhaira also more closely resembles that of thylacosmilines than non-thylacosmiline proborhyaenids. In most borhyaenoids (especially non-thylacosmiline proborhyaenids), the symphysis is broad anteroposteriorly and typically fused in adults. In Eomakhaira, the symphysis is unfused and much narrower, probably ending at the p2/3 embrasure (but perhaps as far posterior as the anterior root of p3). P3 of Eomakhaira is much longer than p3, and the lower premolar row is relatively short. Both conditions are reminiscent of other thylacosmilines, in which dP3 is much longer than p3, and the upper and lower premolar rows are relatively short (though in Patagosmilus and Thylacosmilus, there is a large diastema between the lower canine and premolars due to a more posterior position of the postcanine teeth; Goin, 1997; Forasiepi and Carlini, 2010). The M4 of Eomakhaira resembles that of Patagosmilus more than that of any other borhyaenoid examined in being gracile, anteroposteriorly narrow and labiolingually very wide compared with M1–3 and in having a vestigial protocone. Compared with M4 of Eomakhaira and Patagosmilus, the M4 of Thylacosmilus is anteroposteriorly longer than labiolingually wide (resulting in a robust tooth), due partly to the loss of its talon. Some features common to both Eomakhaira and later thylacosmilines may also occur in Proborhyaena gigantea, especially if MLP 79-XII-18-1 pertains to that species, and potentially represent synapomorphies at a deeper node within Proborhyaenidae; these include a shallow dentary, more posteriorly positioned infraorbital foramen, labiolingually narrow upper canines, and a P3 that is much larger than p3.

Evolution of Saber Teeth in Sparassodonta: Among eutherians, saber teeth have originated independently three or four times: in oxyaenid “creodonts” (Machaeroidinae), in nimravid and barbourofelid carnivoramorphans (if these are not sister taxa that were ancestrally machairodont; Wang et al., 2020), and in felid carnivorans (Machairodontinae) (Antón, 2013). Among metatherians, they are known to have evolved only once, in sparassodonts (suggestions that the extant didelphid Monodelphis dimidiata exhibits a saber-toothed morphology have not been supported by later analyses; Blanco et al., 2013; Chemisquy and Prevosti, 2014). This disparity between eutherians and metatherians is curious given the broad range of predatory metatherians, which include deltatheroidans, stagodontids, the marsupialiform Anatolia-delphys, other groups of sparassodonts (i.e., hathliacynids), didelphids (i.e., sparassocynins, didelphins, and related forms), dasyurids, thylacinids, and thylacoleonids. One potential reason for the rarity of metatherian sabertooth lineages may be related to how their distinctive upper canines may have functioned. It has been argued that saber teeth would have required a considerable learning period to be used effectively (Emerson and Radinsky, 1980; Akersten, 1985; Antón and Galobart, 1999; Wheeler, 2011). Functional modeling of saber bites has shown that an imprecise bite can cause the canines to snag, can easily be be too shallow to be lethal, or can penetrate too deeply to be extracted from the prey (Wheeler, 2011). Elongate, labiolingually narrow upper canines are also vulnerable to breakage when subjected to sudden, unpredictable loads such as those produced by struggling prey (Van Valkenburgh and Ruff, 1987). Many eutherian sabertooths exhibited prolonged retention of the deciduous canines, which are also large and machairodont in these lineages (Bryant, 1988; Wysocki et al., 2015; Wysocki, 2019). It has been hypothesized that this extended retention time resulted in a longer “training” period, in which breakage would not have had permanent consequences (due to their eventual replacement).

Unlike eutherians, metatherians (including sparassodonts; Marshall, 1976c; Forasiepi and Sánchez-Villagra, 2014) lack deciduous canines and therefore have only one tooth generation at the canine locus, precluding “training” canines. At the same time, thylacosmiline sparassodonts are the only sabertooths with hypselodont canines, and this may be a different means to achieve the same end: a tooth that would eventually be replaced in the event of breakage (albeit by growth at the base rather than by wholesale replacement) (Marshall, 1976a). The results of our phylogenetic analysis suggest that open-rooted canines were not a novel innovation of thylacosmilines but a plesiomorphic feature inherited from non-sabertoothed ancestors (i.e., an apomorphy for proborhyaenids). This suggests that the absence of deciduous canines may have constrained the evolution of saber teeth in metatherians and that hypselodonty is a prerequisite for evolving saber teeth in this clade. This may be one reason why saber-toothed canines only evolved once in metatherians rather than repeatedly, as in eutherians. Deciduous canines are widespread within Eutheria, whereas hypselodont canines in metatherians appear to have a much narrower distribution. Therefore, developing saber teeth in metatherians would require not only selection for machairodonty but also the prerequisite of an uncommon morphology.

Open-rooted, ever-growing (hypselodont) canines in proborhyaenids (including thylacosmilines) are frequently considered a unique feature within Metatheria (Vieira and Astúa de Moraes, 2003; Goswami et al., 2011; Forasiepi and Sánchez-Villagra, 2014). However, they may be more widely distributed than generally realized. Hypselodont canines have been reported in peramelemorphians (Aplin et al., 2010; K. Travouillon, personal commun.), and CT scans of adult peramelemorphians show open-rooted canines with nontapering roots and open pulp cavities (Macrini, 2005b, 2007a). A CT scan of the nonsparassodont pucadelphyidan Pucadelphys andinus, interpreted as a sexually mature male (see Ladevèze et al., 2011), appears to show open-rooted upper canines (Macrini, 2007b). The canines of didelphids and dasyuromorphians are reported to be hypselodont by several authors (Jones, 1995, 2003; Voss and Jansa, 2009; Chemisquy and Prevosti, 2014; R. Voss, personal commun.; R. Beck, personal commun.), but we have not been able to confirm these observations. Among the aforementioned studies, Chemisquy and Prevosti (2014) is the only one that provides imagery to support their claim of hypselodonty in Monodelphis dimidiata. However, their radiographs are shown only in palatal view, making it difficult to determine whether the canine pulp cavities are actually open (or merely show the apical foramen). In one of the radiographs said to demonstrate an open root (Chemisquy and Prevosti, 2014: fig. 5B), the canine root is closed. Other CT scans (Macrini, 2001; DigiMorph Staff, 2004; Macrini, 2005a, c) and X-rays (Woolley, 2011; Fiani, 2015) of adult didelphids and dasyuromorphians show nonhypselodont canines. None of the canines of didelphids and dasyuromorphians we have observed exhibit the features associated with hypselodonty that have been described previously in proborhyaenids, such as absence of enamel, remodeling of the apex with well-developed compensatory wear facets, and great height of the tooth despite heavy wear (fig. 20).

The development of hypselodont, open-rooted canines in proborhyaenids may be the result of paedomorphosis. The canine roots of extant marsupials remain open much longer during ontogeny than those of most placentals (Jones, 2003; Chemisquy and Prevosti, 2014), and the canine roots of juvenile non-proborhyaenid sparassodonts remain open relatively late in ontogeny (Forasiepi and Sánchez-Villagra, 2014; Engelman et al., 2015). Combined with the observations above, the presence of closed roots in senescent proborhyaenids, such as SGOPV 3490 and MLP 79-XII-18-1 (Bond and Pascual, 1983), suggests that canine hypselodonty was achieved by delaying root closure until extremely late in ontogeny. Similar evolutionary transitions from closed-rooted to fully hypselodont teeth via postponement of root formation have been observed in other mammals, including notoungulates (Madden, 2015). It has also been suggested that paedomorphosis played a role in other aspects of thylacosmiline evolution, such as the retention of dP3 as a functional element in the adult dentition (Forasiepi and Sánchez-Villagra, 2014).

Other features in thylacosmilines typically related to machairodonty, such as those connected with a wide gape (Emerson and Radinsky, 1980; Slater and Van Valkenburgh, 2008; Antón, 2013), appear to have originated prior to the group's origin (i.e., among non-thylacosmiline proborhyaenids). In Callistoe vincei, the combined heights of the upper and lower canines are comparable to the height of only the upper canine of Thylacosmilus (see Powell et al., 2011). Thus, Callistoe and Thylacosmilus would have required comparable gapes to achieve clearance between the canines (Powell et al., 2011; Wroe et al., 2013).

The transverse processes of the atlas of Callistoe and Thylacosmilus are longer (anteroposteriorly) than wide (transversely), and extend far posterior to the caudal facets. This contrasts with the condition in Prothylacynus, Borhyaena, and Arctodictis, in which these processes are subequal in length and width and extend only slightly posterior to the posterior border of the caudal facets (Argot, 2003; Forasiepi, 2009). Similar distinctions in the morphology of the axial transverse processes are seen in comparisons between saber-toothed and non-saber-toothed felids, respectively (Akersten, 1985; Argot, 2004b; Salesa et al., 2005). Elongate transverse processes in sabertoothed felids increased the area of origin for the obliquus capitis cranialis and obliquus capitis caudalis (Antón and Galobart, 1999), which are thought to be the primary muscles involved in the saber bite in saber-toothed mammals (Akersten, 1985; Antón et al., 2004). The transverse processes of the atlas in Thylacosmilus resemble those of saber-toothed felids (Argot, 2004b: fig. 1C, D). The transverse processes of Callistoe (Argot and Babot, 2011: fig. 2D) extend further posteriorly than those of most sparassodonts but less posteriorly than in Thylacosmilus, suggestive of large obliquus capitis cranialis and obliquus capitis caudalis muscles in this non-saber-toothed taxon. This suggests that specialization of the neck musclature occurred in non-thylacosmiline proborhyaenids prior to the evolution of saber teeth in this clade.

FIG 20.

Lower canines of non-thylacosmiline proborhyaenids. A–B, Proborhyaena gigantea (MACN-A 52-382) in A, labial and B, lingual views. C, Arminiheringia auceta (MACN-A 10970) in labial view. D, Paraborhyaena boliviana (UATF-V-000129; left canine, reversed) in labial view. The upper canines of Arminiheringia show a comparable morphology (see fig. S4B; apices of the upper canines are unknown for Proborhyaena and Paraborhyaena). Scale bars = 10 mm.

The Oligocene Paraborhyaena, which is recovered as more closely related to Thylacosmilus than to Callistoe in our phylogenetic analyses (figs. 16, 17), shows additional similarities to Thylacosmilus in the posterior part of the skull not present in the geologically older Callistoe. (The posterior cranium is unknown in other proborhyaenids, such as Proborhyaena and Patagosmilus.) As in Thylacosmilus, the braincase of Paraborhyaena is short anteroposteriorly (Petter and Hoffstetter, 1983; Muizon et al., 2018), and the nuchal crest is nearly vertical, exposing the occipital condyles in dorsal view (Petter and Hoffstetter, 1983: fig. 2, pl. 4.1B). A shortened temporalis fossa (covarying with an anteroposteriorly short braincase) and vertical occiput, features common among saber-toothed mammals, have been considered to reflect either a wide gape or mechanical compensation of a reduced temporalis muscle lever arm created by a small coronoid process (see Emerson and Radinsky, 1980; Slater and Van Valkenburgh, 2008; Antón, 2013, and references therein). In Proborhyaena, Eomakhaira, and Anachlysictis (in which the temporal and occipital regions of the skull are unknown, but these taxa are phylogenetically bracketed by Paraborhyaena and Thylacosmilus), the coronoid process is large and dorsoventrally tall (Mones and Ubilla, 1978; Goin, 1997), contradicting the latter hypotheses. This suggests that proborhyaenids differed from placental sabertooths in acquiring a short braincase and vertical occiput early in their history (i.e., prior to appearance of saber teeth), whereas these features generally appeared after the acquisition of saber teeth in placentals (Antón, 2013).

Eomakhaira is considered a saber-toothed sparassodont here because it exhibits a degree of machairodont specialization comparable to that of early-diverging members of other sabertoothed clades, e.g., Machaeroides, Dinictis, Nimravus, and Pseudaelurus. Accordingly, sabertoothed sparassodonts (Thylacosmilinae) now have a documented biochron extending from the early Oligocene (32–33 Ma) to the early Pliocene (3 Ma), a duration of almost 30 million years (fig. 21), far longer than that of any placental sabertooth clade. Saber-toothed oxyaenid “creodonts” (Machaeroidinae) span approximately 11 million years (52.8–41.5 Ma; Dawson et al., 1986; Robinson et al., 2004; Kelly et al., 2012; Tomiya, 2013; Zack, 2019a, b), a minimum estimate given the clade's poor fossil record (no more than six species and fewer than 10 specimens). Securely dated and identified nimravids are known from the late Eocene (∼37.8 Ma; Averianov et al., 2016; 2019) to the end of the Oligocene (23 Ma; Bryant, 1996; Peigné, 2003), a temporal range of 14.8 million years. Isolated upper canine fragments from Asia (Chow, 1958; Suyin et al., 1977; Averianov et al., 2016) may push the first appearance datum of this clade back into the late middle Eocene (∼42 Ma), extending its temporal range to approximately 19 million years. However, given that these specimens consist of isolated fragments of upper canine saber teeth, they may pertain to another saber-toothed clade such as machaeroidine “creodonts” (see Zack, 2019b). Barbourofelid carnivoramorphans first appear in the early Miocene (20–19 Ma; Morales et al., 2001; Morlo et al., 2004) and are last recorded in the late Miocene (6 Ma; Tedford et al., 2004), a range of 13–14 million years. Finally, machairodontine felid carnivorans are recorded as early as the middle Miocene (16 Ma) based on the first appearance of Pseudaelurus sensu stricto (Werdelin et al., 2010; Robles et al., 2013) and last appear during the end-Pleistocene extinctions (∼ 0.01 Ma), a temporal range of roughly 16 million years. Excluding Eomakhaira, the biochron of thylacosmilines spans at least 16 million years, based on a Patagosmilus-like upper molar from the early Miocene of Patagonia (Colhuehuapian SALMA, 20.2–20.0 Ma; Goin et al., 2007; Ré et al., 2010). Thus, even excluding Eomakhaira, the temporal range of thylacosmilines exceeds that of nimravids and barbourofelids and is comparable to that of machairodontines.

FIG. 21.

Temporal durations of major lineages of mammalian saber-toothed carnivores, with metatherian lineage in grey and placental lineages in black. A representative skull of each clade is depicted to the right; from top to bottom: Thylacosmilus atrox (Thylacosmilinae), Machaeroides eothen (Machaeroidinae), Hoplophoneus primaevus (Nimravidae), Barbourofelis fricki (Barbourofelidae) and Smilodon fatalis (Machairodontinae). Images of Machaeroides, Barbourofelis, and Smilodon modified from Antón (2013), Thylacosmilus from Riggs (1934), and Hoplophoneus from Scott and Jepsen (1936) and Bryant (1996). Tick mark on Thylacosmilinae record represents the oldest occurrence of this clade (∼20.2 Ma) prior to the discovery of Eomakhaira. Abbreviations: Plio., Pliocene; Ple., Pleistocene.

Sabertooth clades are generally characterized by high rates of extinction and turnover relative to other carnivorous mammals (Naples et al., 2011; Piras et al., 2018). This has been suggested to be related to their inferred specialized hunting behavior and comparatively narrow prey base (based on functional morphology and paleoecological data) and their status as hypercarnivorous apex predators, which would make them more vulnerable to ecological distruptions and environmental perturbations (Naples et al., 2011; Antón, 2013; Piras et al., 2018; and references therein). In this respect, the long stratigraphic range of thylacosmilines compared to placental saber-toothed clades is noteworthy considering the many major faunal/climatic changes faced by the former in South America, including the Bisagra Patagónica during the Oligocene (Goin et al., 2010), the faunal turnover at the Oligocene-Miocene boundary (which, notably, marked the end of all non-thylacosmiline proborhyaenids; Bond and Pascual, 1983), the Middle Miocene Climatic Optimum and subsequent climatic deterioration (Croft et al., 2016), and the expansion of grasslands during the late Miocene (Pascual and Ortiz Jaureguizar, 1990). The shorter stratigraphic ranges of placental sabertooths relative to thylacosmilines do not appear to be attributable to competitive interactions between placental sabertooth clades in North America, Eurasia, and Africa compared to the relative isolation of thylacosmilines in South America, as placental sabertooth clades mostly do not overlap in space and time (e.g., North America's “cat gap” of Hunt and Joeckel, 1988). The large disparity between the temporal ranges of thylacosmiline metatherians and placental sabertooth clades suggests dissimilar ecological requirements, with thylacosmilines perhaps having broader dietary and/ or habitat preferences. Thylacosmiline sparassodonts, a distinctive, diverse, and temporally long-lived lineage of machairodont mammals, hardly represent “inferior” or “ineffective” imitations of their placental analogs as frequently claimed (Riggs, 1934; Simpson, 1940; Patterson and Pascual, 1972; Werdelin, 1987; McNab, 2005; Prothero, 2006; Webb, 2006; Leigh et al., 2014; Faurby and Svenning, 2016; Faurby et al., In press). The discovery of Eomakhaira clarifies the evolution of machairodonty both within Sparassodonta and in mammals generally and speaks to the enduring utility of the South American fossil record in elucidating the splendid variety of morphological diversity among mammals and nature's fantastic capacity for convergence.