Anachlysictis gracilis Goin, 1997

Suarez, Catalina, Forasiepi, Analia M., Babot, María Judith, Shinmura, Tatsuya, Luque, Javier, Vanegas, Rubén D., Cadena, Edwin A. & Goin, Francisco J., 2023, A sabre-tooth predator from the Neotropics: Cranial morphology of Anachlysictis gracilis Goin, 1997 (Metatheria, Thylacosmilidae), based on new specimens from La Venta (Middle Miocene, Colombia), Geodiversitas 45 (18), pp. 497-572 : 503-520

Anachlysictis gracilis Goin, 1997

( Figs 2-11 ; 12C )

HOLOTYPE. — IGM 184247 ( Fig. 2 ), a small portion of skull roof with the left postorbital process, and fragments of frontal, lacrimal and nasal; nearly complete right mandible with almost complete m2-4 and roots of p2-m1; left horizontal ramus fragment with m2-3; fragment of left symphyseal flange; almost complete atlas; fragment of the third cervical vertebra; fragments of neural arc, pre and post-zygapophyses of undetermined vertebral elements; ribs fragments; proximal portion of right scapula; right magnum; pyramidal?; two distal fragments of metapodials; proximal phalanx; and indeterminate postcranial fragments.

REFERRED SPECIMENS. — UCMP 39705 ( Fig. 3 ), a posterior fragment of the right mandibular ramus, preserving the condyle and angular process; VPPLT-1612 ( Figs 4-11 ; see also Appendices 3-5), a partial skeleton, including its nearly complete skull.

LOCALITIES AND STRATIGRAPHY. — IGM 184247 (holotype), IGM locality 75, level between the Chunchullo Sandstone Beds and the Tatacoa Sandstone Beds ; UCMP 39705 , UCMP locality V4531 (“ Cerro Gordo 2”), level between the Chunchullo Sandstone Beds and the Tatacoa Sandstone Beds ; VPPLT-1612 , locality “ Finca Tres Pasos ”, La Victoria, Chunchullo Sandstone Beds (StL 4 in Mora-Rojas et al. 2023). These localities are part of the La Venta area, La Tatacoa Desert, Huila Department, Colombia; La Victoria Formation, Honda Group, Middle Miocene, Laventan SALMA .

EMENDED DIAGNOSIS. — Differs from other thylacosmilids in having a proportionally lower and longer skull (longer than twice its width at the level of the zygomatic arch); facial and dorsal portion of the skull flatter; posterior end of palate concave, single arched; mandible much slender and with the symphyseal flange less developed; upper canine not triangular in cross-section and with much shorter root whose implantation is less dorsalized; postcanine teeth rows (upper and lower) less bowed; and protocones and talonids more developed (modified from Goin 1997: 202). Differing from P. goini in a less developed postorbital process; less developed juga alveolaria; upper canine proportionally wider (less laterally compressed); P2 single-rooted.

MEASUREMENTS. — See Table 1.

COMMENT

Several characters from the diagnosis proposed byGoin (1997) were excluded from the emended diagnosis, as they are now known to be also present in P. goini .

DESCRIPTION

The description here proposed is based on the holotype and the new referred specimens reported in this work.

Skull

The skull of A. gracilis is relatively gracile compared with that of T. atrox . It is relatively lower and longer than the skull of T. atrox and P. goini ( Fig. 12 ). The snout is longer than wide ( Fig. 4 ) and lower than in T. atrox and P. goini ( Fig. 12 ). The rostrum of A. gracilis (measured from the anterior margin of the orbit and the anterior end of the snout (approximately) is longer than T. atrox and P. goini or even longer than in borhyaenids.

The maximum width of the skull is at the level of the cranial vault (between the roots of the zygomatic arches), being c. 30% wider than the postorbital widest point, differing from T. atrox , in which the maximum width is at the level of the postorbital bars. The postorbital constriction is wellmarked. The braincase is c. 30% longer than wide. In lateral view ( Figs 6 ; 7 ), the dorsal outline of the skull is roughly straight, slightly curving upwards at the level of the nasofrontal suture and descending forward in a convex outline. This shape is notably different from the skull of T. atrox , where the dorsal outline is curved, being dorsally convex in almost all its extension (except at the level of the cranial vault). The occipital condyles project posteriorly approximately until the level of the nuchal crest, differing from the condition in T. atrox , where they protrude posteriorly markedly beyond. Breakage and deformation preclude many comparisons with the skull shape of P. goini .

The palate is at the same level as the basicranium floor (Appendix 3). In ventral view ( Fig. 5 ), the palate is roughly triangular, diverging backwards and reaching the maximum width at the level of M3. The lateral edge of the palate, corresponding to part of the maxilla, is markedly high, as in P. goini ( Forasiepi & Carlini 2010) and T. atrox ( Riggs 1933, 1934; Goin & Pascual 1987), and the postcanine teeth are set in concave (bowed) arcades, less marked than in P. goini and much less than in T. atrox . In more generalized sparassodonts, the postcanine upper tooth row is generally laterally straight or nearly straight ( Sinclair 1906; Babot et al. 2002; Forasiepi 2009).

The premaxilla is poorly preserved, cracked, and partially broken. In palatal view, the sutures with the maxilla are obscured by breakage, and it is not possible to see the posterior extension of the lateral palatal process of the premaxilla ( Fig. 5 ). The bone around the incisive foramina was poorly preserved, so their posterior limits are unclear. However, we assume that the bone around the foramen corresponds to the premaxilla anteriorly and laterally, and the maxilla posteriorly, as in other sparassodonts and marsupials (e.g., Sinclair 1906; Wible 2003; Babot et al. 2002; Forasiepi 2009; Forasiepi et al. 2015). There is a thin medial bridge of bone between the incisive foramina, which is part of the medial palatal process of the premaxilla. As suggested by the preserved portion, this structure would be placed in a horizontal position. The foramen occupied a position mostly anterior to the canine with a small extension between the canines (the posterior margin of the incisive foramina is posterior to the anterior edge of the canines). In dorsal view, the posterior end of the facial process of the premaxilla (the posteriormost point of the premaxilla-nasal contact) projected backwards, reaching the level of the upper canine (posterior to its anterior edge; Figs 4 ; 7 ). The paracanine fossa is deep and delimited by a crest (it is better visible in ventral view: Fig. 5 ) and there is not an evident precanine notch. However, the lateral wall of this fossa is crushed, and it is not possible to identify if it is formed only by the premaxilla, as seen in all the sparassodonts with this portion preserved (e.g., Acyon myctoderos Forasiepi, Sánchez-Villagra, Goin, Takai, Shigehara & Kay, 2006 , Cladosictis patagonica Ameghino, 1887 , Lycopsis longirostrus Marshall, 1976 , Prothylacynus patagonicus Ameghino, 1891 , Thylacosmilus atrox , borhyaenids, etc.), or if there is participation of the maxilla.

The maxilla is exposed in dorsal, lateral, and palatal views, and on the orbit floor (in dorsal and dorsolateral views; Figs 4 ; 6B, C ). This bone contributes significantly to the lateral aspect of the skull (though less than in T. atrox ). It is separated from the frontal by the nasal and lacrimal, which are in contact, as in other sparassodonts ( Figs 6 ; 7 ), including P. goini (e.g., Sinclair 1906; Petter & Hoffstetter 1983; Marshall 1976; Babot et al. 2002; Forasiepi & Carlini 2010; Forasiepi et al. 2015). However, it differs from the condition of T. atrox , in which the maxilla projects dorsally over the dorsal surface of the skull and posteriorly beyond the level of the orbit (forming an ascendant dorsally convex surface), in company of the hypertrophied root of the ever-growing canine ( Gaillard et al. 2023); its posterior border has an extended contact with the frontal bones.

In lateral view ( Figs 6 ; 7 ), the canine root defines a well-developed swelling (corresponding to the juga alveolaria), which projects dorsally, reaching the suture with the nasal. However, this swelling seems to be less prominent than in P. goini (but the skull is distorted and this feature seem to be artificially stressed; Fig. 12 ). In T. atrox , this is much more stressed; see the Discussion), however it looks different because this swelling follows all the long dorsal projection of the bone extending posteriorly to the orbit. The infraorbital foramen is relatively small compared to that of other sparassodonts (e.g., Borhyaena Ameghino, 1887 ; see Sinclair 1906), it is placed located dorsal to the posterior root of the P3, as in P. goini and T. atrox . There is neither an anteorbital fossa nor a small foramen on the facial aspect of the maxillary below the infraorbital foramen (near the alveolar border), like in T. atrox ( Riggs 1934: 10, pl. I; Goin & Pascual 1987). Below the zygomatic arch, the maxilla bears a small shallow depression at the level of the M3 (more clearly observable on the right side; Fig. 7 ) that probably corresponds to the depression for the masseter muscle ( Turnbull 1970).

In ventral view, the maxilla is not expanded behind the infraorbital foramen as in several sparassodonts (e.g., Cladosictis Ameghino, 1887 , Arminiheringia Ameghino, 1902 , Callistoe Babot, Powell & Muizon, 2022 , Borhyaena , Arctodictis Mercerat, 1891 ) where the maxilla is markedly flares forming “cheeks” ( Marshall 1981; Babot et al. 2002; Forasiepi 2009). The condition in A. gracilis is closer to that of P. goini and T. atrox , with almost flat maxilla (flatter than in A. gracilis ).

On the palatal process of the maxilla, there are palatal pits between all upper molars. These circular depressions (for the reception of the protoconids when the jaws are closed) are deeper between the M2-3 and M3-4 ( Fig. 5 ). The palatal surface of the maxilla shows several minute foramina (to transmit the major palatine nerve and accompanying vasculature: Forasiepi 2009, characters 22 and 23), as in other sparassodonts (e.g., Sinclair 1906; Marshall 1976; Babot et al. 2002; Forasiepi & Carlini 2010; Forasiepi et al. 2015), instead of having large fenestrae (vacuities) as seen in living marsupials, or individual major palatine foramina as in placentals. A pair of small minor palatine foramina, nearly circular, is present lateral to both sides of the choanae edges at the level of the posterior-most palatal pit (between the M3 and M4; Fig. 5 ). They are placed on the maxillo-palatine suture and are largely formed by the palatine and, in less proportion, by the maxilla (contributing to the lateral margin). This position is similar to that of T. atrox ( Riggs 1934: 19, pl. II-2) and the condition inferred by Forasiepi & Carlini (2010) for P. goini . However, in at least one specimen of T. atrox (MLP 35-X-41- 1), the minor palatine foramen opens entirely in the palatine bone ( Forasiepi & Carlini 2010).

A pair of well-defined semicircular notches is present posterior to the minor palatine foramina (at the level of the posterior margin of the M4), medially to each tooth row (see “n” in Fig. 5B ). Similar notches (in shape and location) have been identified in some non-thylacosmilid sparassodonts (e.g., Arctodictis sinclairi Marshall, 1978 ; see Forasiepi 2009). They have been interpreted as corresponding to the anterior margin of the minor palatine foramina, which would be opened posteriorly, lacking its posterior bridge (see Forasiepi 2009: character 24). However, in the specimen VPPLT-1612, it is possible to see this structure, while a minor palatine foramen (well-defined, with closed margins) is also clearly identifiable (see “mpf” in Fig. 5 ).

The zygomatic arch is formed by the jugal and squamosal. It is longer and slenderer than in T. atrox ( Riggs 1933, 1934) but similar to other sparassodonts, such as the borhyaenoids (e.g., Sinclair1906; Forasiepi 2009; Forasiepi et al. 2015). The maxillajugal suture is irregular and forms a roughly zigzag line, as in T. atrox ( Fig. 7 ). The suture between the jugal and squamosal is almost straight. In lateral view ( Fig. 4 ), the anteriormost edge of the jugal reaches the level of the M2 (approximately posterior to the metacone). The external surface of the zygomatic arch (including both jugal and squamosal) has long, narrow parallel striae. The anterodorsal margin of the jugal is concave, forming the ventral border of the orbit and ending posteriorly in a frontal process of the jugal, which is tall. The jugal usually ends posteriorly bifurcated in two branches, one dorsal and one ventral. Only the ventral branch is completely preserved in the specimen VPPLT-1612 ( Fig. 7 ). It is ventrally curved and extends posteriorly to form the preglenoid process.

The palatine (paired) contributes to the posterior hard palate, the nasopharyngeal passage, and the medial wall and floor of the orbit. In ventral view, the palatine contacts the maxilla anteriorly and the presphenoid and pterygoid posteriorly in the nasopharyngeal passage ( Fig. 5 ). In this view, the palatines extend anteriorly until the level of the M2, forming an irregular parabolic suture with the maxilla; and posteriorly to the level of the last molar, as in P. goini and Borhyaena tuberata Ameghino, 1887 . The posterior end of the palatines on the palate is slightly thicker than the rest of the horizontal plate, forming the border of the choanae. This border is posteriorly concave, single-arched, differing from T. atrox , with a double-arched margin. The choanae open at the level of the contact between M3-4. The portion of the palatines exposed into the nasopharyngeal passage (immediately behind the choanae border) is strongly cracked. However, it is possible to see that they are well developed and expanded on the medial side but without midline contact, as the presphenoid is visible between them ( Fig. 5 ).

In dorsal and dorsolateral views, the palatine is exposed on the orbit floor and separated from the maxilla by an irregular suture. It contributes to the infraorbital canal. In lateral view, the sutures delimiting the palatine are interrupted by bone fractures. However, it is possible to see the contact with the maxilla and lacrimal anteriorly and the frontal dorsally. The palatine also contacts the orbitosphenoid, alisphenoid, and pterygoid posteriorly, but these sutures are difficult to differentiate due to the specimen’s poor preservation (see Fig. 6 ). At the junction of the floor and lateral wall of the orbit, the sphenopalatine foramen is observed, though poorly preserved. This aperture is close to the anterior border of the orbit, as in P. goini ( Forasiepi & Carlini 2010) , while in other sparassodonts, including T. atrox , this aperture is more posterior ( Riggs 1934; Forasiepi 2009; Forasiepi & Carlini 2010).

The pterygoid is a paired bone exposed in the mesocranium in ventral and lateral views. In ventral view, the pterygoids are well developed and expanded on the medial side but without middle contact, exposing the presphenoid. Posteriorly, the pterygoids project posterolaterally (at each side) on the ventral surface of the alisphenoid, forming thin ribbons (i.e., probably part of the pterygoid hamulus or hamular processes), better preserved on the right side ( Figs 5 ; 8 AB).

The nasals are well exposed in dorsal view ( Fig. 4 ), they are very narrow anteriorly and broad posteriorly, widening abruptly at the level of the orbit, as in P. goini ( Forasiepi & Carlini 2010) . This difference in width is even more conspicuous than in other sparassodonts ( Sinclair 1906; Babot et al. 2002; Forasiepi 2009). This morphology differs from T. atrox , where the nasals narrow backwards in dorsal view, although they are long ( Riggs 1934; Turnbull & Segall 1984), compressed between the maxilla and reaching the level of the orbit. The nasofrontal suture is posteriorly convex, forming an open “U”, as in P. goini ( Forasiepi & Carlini 2010) , similar to other sparassodonts ( Sinclair 1906; Forasiepi 2009) and differing from T. atrox , where the frontal is not in contact with the nasal, because of the enlargement of the maxilla as described before (e.g., Riggs 1934; Marshall 1976; Goin & Pascual 1987; Muizon 1999).

In lateral view ( Figs 6 ; 7 ), the lacrimal of A. gracilis extends anteriorly beyond the orbit and orbital rim, with a relatively wide exposition on the rostrum (the width of the facial process of the lacrimal is more than half of its height; Fig. 4 ), similar to the condition seen in T. atrox and some non-thylacosmilid sparassodonts with the lacrimal extended onto rostrum (e.g., Callistoe vincei , Arctodictis sinclairi , Borhyaena tuberata , Prothylacynus patagonicus ). There is one lacrimal foramen on each lacrimal bone, inside the orbit, a generalized condition among sparassodonts ( Sinclair 1906; Riggs 1934). The lacrimal tubercle in A. gracilis is less developed than in P. goini and T. atrox (relatively more developed in the latter).

The frontals contact the nasals and lacrimals anteriorly, the palatine and alisphenoid ventrally, and the parietals and squamosal posteriorly, all by irregular sutures. The suture with the nasals is posteriorly convex (seen in dorsal view: Fig. 4 ); the one with the lacrimal and palatine is roughly anteriorly convex (visible in orbital view; Fig. 6B, C ). The suture with the parietals is transverse (visible in dorsal view: Fig. 4 ); while the one with the squamosal is slightly anteriorly convex (visible in lateral view: Fig. 7 ). The point of contact between the nasal, lacrimal, and frontal is located approximately at the level of the postorbital process in the holotype of A. gracilis ( Fig. 2D ), as in P. goini (the appearance of this portion of the skull is almost identical). However, the condition in the specimen VPPLT-1612 is different, with the contact anterior to the postorbital process ( Fig. 7 ), similar to other sparassodonts (e.g., Hondadelphys Marshall, 1976 , Sallacyon Villarroel & Marshall, 1982 , Acyon Ameghino, 1887 , Cladosictis , Prothylacynus Ameghino, 1891 , Arctodictis , Pharsophorus Ameghino, 1897 , Callistoe ). In T. atrox , the relationship between these bones is different because the posterodorsal projection of the maxilla interposes between the nasal and lacrimal. However, the posteriormost point of the lacrimal-frontal suture extends beyond the level of the postorbital bar, being even more posterior than in A. gracilis and P. goini . Besides these three thylacosmilid taxa (excepting the specimen VPPLT-1612), none of the other metatherian taxa observed for the present work (see Material examined in Appendix 1) shows the lacrimal reaching the level of the postorbital process (or bar).

The postorbital processes are well developed ( Figs 4 ; 6 ), similar to P. goini , and differing from T. atrox , where there is a bony postorbital bar, being the only sparassodont with the orbit completely separated from the temporal fossa ( Riggs 1933, 1934), and one of the few metatherians with a complete osseous postorbital bar ( Gaillard et al. 2023). A small foramen is present anteroventrally to the postorbital process, probably corresponding to the foramen for the frontal diploic vein ( Fig. 7 ), following a similar structure recognized in Monodelphis domestica Wagner, 1842 (see Wible 2003). The temporal lines are weak and contact at the mid-line of the skull, forming the anterior base of the sagittal crest ( Figs 4 ; 6 ). This condition is similar to P. goini ( Forasiepi & Carlini 2010) , resembling other sparassodonts and differing from T. atrox , with temporal lines strongly developed, converging more posteriorly in the skull and describing a sigmoid line ( Riggs 1933, 1934; Forasiepi & Carlini 2010).

The parietal is paired and both elements form most of the roof of the skull ( Fig. 4 ). It contacts the frontals anteroventrally, the squamosal posteroventrally, and the interparietal posteriorly, by irregular sutures. The suture between the parietals and frontals shows a posterior wedge of frontals entering between the parietals. The contact with the interparietal is partially visible in VPPLT-1612; in P. goini (see referred specimen: Material and methods, and Appendix 1, Material examined) the suture is incomplete due to partial fusion with the parietal. In T. atrox , the interparietal is also distinguishable from the parietal, as in several metatherians (e.g., Didelphis Linnaeus, 1758 , Monodelphis Burnett, 1830 , Macropus Shaw, 1790 , Sipalocyon Ameghino, 1887 , and Cladosictis ; Clark & Smith 1993; Forasiepi 2009), in contrast with other borhyaenoids, where it is not (e.g., Arctodictis , Lycopsis Cabrera, 1927 , Borhyaena ; seeForasiepi 2009). Fusion with parietals is inferred in those cases in which suture is not observed, since presence of interparietal seems to be primitive common pattern of mammals ( Koyabu et al. 2012). The interparietal contacts the squamosal anteroventrally by an irregular suture and forms most of the nuchal crest.

The sagittal crest is formed at the midline of the skull by frontals, parietals and interparietals and becomes taller posteriorly ( Figs 4 ; 6 ; 7 ; see also Appendix 3). It is well developed and long, extending from the anterior area of the temporal fossa to the nuchal crest, similar to P. goini ( Forasiepi & Carlini 2010) ; in T. atrox , the sagittal crest is considerably shorter and much more taller and robust (likely to enlarge the area of attachment of the temporal muscle by considering the shorter length of the temporal fossa; Gaillard et al. 2023). The nuchal crest is located at the posterodorsal border of the skull ( Figs 6 ; 7 ; Appendix 3). It is formed by the interparietal and supraoccipital medially and the squamosal lateroventrally. It is well developed, flaring posterolaterally and posterodorsally, and extends back to the level of the occipital condyles (see Appendix 3), as in P. goini , but differing from T. atrox , where the nuchal crest is located more anteriorly, thus the condyle is fully visible in a dorsal view of the skull.

The squamosal forms the posterior portion of the zygomatic arch and the posteroventral portion of the temporal region; it also contributes to the walls of the middle ear cavity as seen in ventral view. It contacts the frontal anteriorly, the alisphenoid anteroventrally, the jugal anteriorly and ventrally (in the zygomatic arch), the frontal anterodorsally, the parietal dorsally, and the interparietal posteriorly ( Figs 4-8 ). In ventral view, the squamosal contacts the jugal at the anterolateral margin of the glenoid cavity and the alisphenoid medially in the ear region ( Fig. 5 ). The zygomatic process of the squamosal, in lateral view, has a roughly elongated shape with well-defined borders. In lateral view, the temporal portion of the squamosal is well developed, roughly semicircular, and its surface is convex, defined by an irregular suture ( Figs 4 ; 6 ; 7 ). The suprameatal foramen opens on the posterolateral region of the squamosal ( Figs 6 ; 7 ; Appendix 3), at the level of the external acoustic meatus and above the suprameatal crest. This foramen is oval-shaped and opens posterodorsally.

The glenoid cavity is formed only by the squamosal, because the preglenoid process of the jugal contacts it but does not contribute to the cavity, lacking an articular facet (as in P. goini and T. atrox ). The alisphenoid does not either contribute to the glenoid cavity ( Fig. 5 ) and there is no alisphenoid glenoid process, as in marsupials (see Wible 2003). The glenoid cavity is concave and ellipsoidal, with the transverse length less than twice the anteroposterior width and faces ventrally ( Fig. 5 ). In lateral view, the preglenoid process of jugal and the postglenoid process of the squamosal have a similar ventral extension (better seen on the right side, with less deformation), though the postglenoid process is wider (lateromedially) and relatively more robust anteroposteriorly than the preglenoid process. The postglenoid foramen is placed on the anterior wall of the external acoustic meatus, medial to the postglenoid process.

Regarding the sphenoid complex, the orbitosphenoid and alisphenoid are exposed in the lateral wall, while the presphenoid and basisphenoid are on the skull floor. However, the synchondrosis between the components of the sphenoid complex are only partly seen. In lateral view, the orbitosphenoid is small, in contact with the alisphenoid, palatine, and frontal. The sphenorbital fissure is the largest aperture on the lateral wall of the skull. It is limited by the orbitosphenoid anteriorly and the alisphenoid posteriorly ( Figs 6B, C ; Appendix 3) as in other metatherians (e.g., Wible 2003).

The alisphenoid mainly contributes to the lateral wall of the braincase and the anterior wall of the tympanic cavity. In lateral view, the alisphenoid contacts the pterygoid, palatine, and orbitosphenoid anteriorly, the frontal dorsally, and the squamosal posteriorly ( Fig. 6B, C ). On the alisphenoid, the foramen rotundum is located posterior to the sphenorbital fissure, opening at the angle between the lateral wall and the floor of the infratemporal fossa. The foramen rotundum is small compared to the sphenorbital fissure, round, and anteriorly directed. There is a relatively wide and shallow sulcus running forward from the foramen rotundum ( Fig. 6B, C ), which was likely occupied in life by the maxillary division of the trigeminal nerve V 2 and accompanying vessels ( Sisson 1965; Hiatt 2020).

In ventral view, there is no contribution of the alisphenoid to the tympanic floor (it lacks an alisphenoid tympanic process similar to T. atrox ; Forasiepi et al. 2019). The foramen ovale is bounded by the alisphenoid only (without any contribution of the petrosal). Similar to T. atrox and the specimen from Quebrada Honda referred to P. goini (see Material and methods, andAppendix 1, Material examined), there is a posterior rod of alisphenoid bone that posteriorly limits the foramen ovale as seen in the right side of the skull. This piece of bone does not qualify as a component of the alisphenoid tympanic process because it does not participate in bounding the middle ear cavity and for that reason, we interpret that there is not a secondary foramen ovale, like in other sparassodonts (see Forasiepi 2009).

The middle ear cavity is so poorly preserved and fragile that its preparation could not be completed. In consequence, neither the petrosal (which we do not discard it could be preserved internally, as in Thylacosmilus ; Forasiepi et al. 2019) nor the suture with this bone is observable by surface examination; similarly, its suture with the exoccipital is not clearly visible.

In ventral view, the presphenoid, basisphenoid and basioccipital forms the floor of the caudal cranium. The synchondroses between sphenoid elements are not clear. The presphenoid exposes on the nasopharyngeal passage. The basisphenoid is roughly triangular and anteroposteriorly elongated ( Fig. 5 ). It contacts the basioccipital posteriorly, through an almost transverse synchondrosis, and the alisphenoid laterally. There are two thick, robust, and parallel crests, the sphenoidal tubercles or basilar tubercles ( Riggs 1934; Forasiepi et al. 2019), which decrease in height anteriorly ( Fig. 8A, B ; more details in Appendix 3). These structures are much weaker than those of T. atrox , where they are hyperdeveloped. In P. goini , they are intermediate in size. Lateral to the basilar tubercles, there is a very narrow and shallow groove that becomes shallower anteriorly ( Figs 5 ; 8A, B ; Appendix 3). The carotid foramen opens posterolaterally anterior to the basisphenoid-basioccipital synchondrosis is ( Fig. 8A, B ).

The posterior floor of the caudal cranium is formed by the basioccipital. It contacts the basisphenoid anteriorly and the exoccipitals posterolaterally. It would also contact the petrosal laterally as in other marsupials and sparassodonts (e.g., Babot et al. 2002; Wible 2003; Forasiepi 2009; Forasiepi et al. 2019), but matrix prevents further confirmation. In ventral view, the basioccipital is roughly rectangular (with the lateral borders laterally convex; Figs 5 ; 8A, B ). In ventral view, and at the anterolateral limit with the condyle, there is a large foramen; posterior to it, there is another one, similar in size but placed on the condyle surface ( Fig. 8A, B ). Both foramina open anteriorly and correspond to two hypoglossal foramina (rostral and caudal). At each side of the basioccipital, there is a deep basijugular sulcus ( Forasiepi et al. 2019) running with a constant width in direction to the rostral hypoglossal foramen ( Fig. 8 ). This sulcus ends anteriorly in a concavity that could correspond to the foramen for the inferior petrosal sinus but its margins are broken ( Fig. 8 ). For the same reason, the morphology of the jugular foramen (or fossa) and its convergence with the inferior petrosal sinus could not be evaluated for this specimen.

In posterior view ( Fig. 8C ), synchondroses between the occipital bones (or their absence due to fusion) are not seen due to the bad preservation condition of the occiput. Similarly, it is also unclear if the mastoid portion of the petrosal contributes to the occipital shield, and the extension of the contribution of the squamosal. The dorsal portion of the occipital shield is slightly concave in posterior view. It contacts the interparietal dorsally at the level of the nuchal crest. Numerous minor grooves and rugosities are mainly distributed on the lateral sides of the occipital surface ( Fig. 8C ). All these scars would correspond to the attachment area of the nuchal musculature ( Turnbull 1970). There are also several small foramina, likely related to the feed of the occipital musculature. The occipital condyle protrudes posteriorly and is ellipsoidal in posterior view, with the longer axis in a horizontal position. The articular facets of the condyle continue at the sagittal plane, connecting ventrally.

Dentary

The dentary is intermediate in height (depth below m3/m4 embrasure/total length of dentary = 0.17), and the horizontal ramus represents approximately 60% of the total length. The ventral border of the dentary is nearly horizontal behind the level of the p2 ( Figs 2A ; 9 ; Appendices 4, 5). Anterior to the p2, the dentary increases in height and its ventral border curves, expanding ventrally, forming a symphyseal flange ( Figs 2A, C ; 9 ; Appendices 4, 5), less developed than in T. atrox and P. goini (inferred from a portion present in the specimen from Quebrada Honda; see Material and methods, and Appendix 1, Material examined). The dentary is laterally concave at the level of the canine-premolar series and laterally convex at the level of the molar series, resulting in a sigmoid morphology, better seen in dorsal and ventral views ( Figs 2B ; 9C ; Appendices 4, 5), less marked than in P. goini and much less than in T. atrox . This morphology is accompanied by a bowed lower postcanine tooth row, slightly less marked than in the upper arcade and less than in the other thylacosmilids.

The height of the coronoid process is at least twice that of the horizontal ramus (behind the p2), being taller than in T. atrox , where it is also strongly reduced. The angle between the anterior coronoid crest and the alveolar border is c. 116°, greater than in T. atrox (c. 90°). The masseteric crest is horizontal at the three quarters anterior while its posteriormost portion bends upwards, forming an obtuse angle, and projects along the coronoid process as a continuous, flat shelf. The masseteric fossa is wide and well developed, differing from the strongly reduced fossa in T. atrox . The mandibular condyle is cylindrical, oval in posterior view, with a well-defined mandibular neck, and located at the level of the tooth row as in T. atrox ( Figs 3 ; 9 ; Appendix 5). The shape of the angular process is shelf-like (followingSánchez-Villagra & Smith 1997) and extends medially slightly beyond the level of the medial end of the mandibular condyle (Appendix 5).

In lateral view, three mental foramina are present: the anteriormost and largest is located on the symphyseal flange slightly posterior to the level of the canine; the second is located below the p2-3 contact level; and the third, below the anterior root of the m1 ( Figs 2A, C ; 9B ). The specimen VPPLT-1612 shows what apparently could be a broken fourth mental foramen below the anterior root of the m4. However, it is not clear if it is, in fact, a foramen or just an artificial aperture, because the bone is broken at this exact point in both the left and right dentaries. The anteriormost mental foramen in A. gracilis is located on the lower portion of the symphyseal flange, being topographically not aligned with the other foramina but markedly lower, surpassing ventrally the level of the ventral margin of the horizontal ramus (the portion behind the flange), similar to T. atrox (in this species the foramen is below the lower canine). This condition differs from that in other compared sparassodonts in which the first mental foramen is aligned or nearly aligned to the other foramina. In medial view, the mandibular foramen of A. gracilis is large, located at the midpoint of the coronoid process in the holotype, similar to T. atrox , and posterior to this point in the specimen VPPLT-1612 ( Fig. 9A ).

Dentition

The dental formula of A. gracilis is: I4?/i?, C1/c1, P2/p2, M4/ m4. The premolars are uninflated and reduced in number (see below). The dentition shows a marked carnassial morphology: a moderate reduction of the protocone, talonid basin (participants in the crushing mechanism), and stylar shelf; and a strong development of cutting blades (preparacrista, postmetacrista, and paracristid). However, these carnivorous adaptations are slightly less developed than in P. goini and much less than in T. atrox (and other sparassodonts such as borhyaenids). Although some cusps are secondarily broken, it is possible to see a moderate wear (in both specimens with dentition preserved: the holotype and VPPLT-1612). Additionally, the molars are completely erupted ( Figs 10-11 ; Appendix 3), indicative of an adult or young adult ontogenetic stage.

Upper dentition. The premaxilla is broken anteriorly in the specimen VPPLT-1612 (not preserved in the holotype), but there is evidence of at least three incisors at the right side: an incomplete alveolus with a fragment of root, followed by a second alveolus preserving the root, and a posterior incisor with almost complete crown (only the tip is broken; Fig. 10E ). However, although the bone is broken at the anterior margin of the palate, there is enough space between the first preserved alveolus and the midline to allocate an additional incisor, suggesting the presence of four incisors in this specimen. Following the position of the incisors on the right side, the arcade would be slightly parabolic in shape. The upper incisors, if present, are unknown in P. goini and T. atrox but based on wear surfaces on the lower incisors and the transverse breadth between the upper canines, T. atrox could have possessed upper incisors ( Churcher 1985; Goin & Pascual 1987). However, no specimen has yet been collected preserving upper incisors.

Such as T. atrox and P. goini , the most outstanding feature of the upper dentition of A. gracilis is the hyperdeveloped canine. This tooth is long, narrow, and sabre-like. It is proportionally less developed than in P. goini , being vertically and anteroposteriorly shorter, though relatively slightly wider (laterally less compressed), and markedly less developed than in Thylacosmilus . In lateral view, the anterior border is convex and more curved, while the posterior is also convex at the proximal portion but concave at the distal (Appendix 3). The anterior border is blunt and slightly thicker, whereas the posterior one is sharp ( Fig. 7 ), forming a crest which is covered by enamel. This enamel is slightly more extended on the labial side than on the lingual and is mostly restricted to the posterior crest, but it widens distally, extending on the entire tip surface. This condition is partially similar to that in P. goini , but in that species the enamel is more extended on the labial surface; and differs from T. atrox , with the enamel extended along the entire tooth. The surface of the canine root has wide and shallow longitudinal sulci: three shallower (almost vestigial) on the labial face and one, wider and deeper, on the lingual. Additionally, there are small and shallow ridges (wrinkles) restricted to the base of the canine. The labial surface of the canine is slightly convex, differing from P. goini and T. atrox , where this surface is divided into two facets, forming a triangular shape in transverse section (being more marked in the last).

Only two premolars are present in the upper dental series. The first (serially homologous to the P2, as interpreted for the thylacosmilids; see Forasiepi 2009; Forasiepi & Carlini 2010; Suarez 2019) is represented only by a root fragment at the right side and an alveolus at the left ( Fig. 10 A-D). The preserved portion shows a strongly reduced tooth, apparently single-rooted, preceded and followed by long diastemata and set nearly equidistantly separated from the canine and the following premolar ( Fig. 10A, B ), similar to P.goini ( Forasiepi & Carlini 2010) . In T. atrox , the P2 is also single-rooted but set closer to other cheek teeth than the canine ( Riggs 1933, 1934; Marshall 1978a; Goin & Pascual 1987; Forasiepi & Carlini 2010). On the other hand, in P.goini , this tooth was apparently double-rooted and nearly equidistantly separated from the canine and the following premolar ( Forasiepi & Carlini 2010). The last premolar of the specimen VPPLT-1612 (P3) is almost complete on the left side of the skull, and it preserves only the roots on the right side ( Fig. 10 A-D). It is a small molariform tooth, as in the P3 of P. goini , resembling the deciduous premolar of other sparassodonts ( Sinclair 1906; Marshall 1978a; Forasiepi & Carlini 2010; Forasiepi & Sánchez-Villagra 2014), and has three roots (one lingual and two labial). A third upper molariform premolar with three roots has been previously described for T. atrox ( Riggs 1933; Goin & Pascual 1987; Mones & Rinderknecht 2004; Forasiepi & Carlini 2010) and interpreted as a retained deciduous tooth DP 3 in the adult dentition ( Goin & Pascual 1987; Forasiepi & Carlini 2010; Forasiepi & Sánchez-Villagra 2014). Due to the wear degree in the specimens studied here (indicating adult individuals), we extend this interpretation to A. gracilis .

The upper molars increase in width posteriorly, and the length increases from the M1 to the M3, while the M4 is anteroposteriorly short (because it lacks the metastylar region; Fig. 10 A-D). The paracone and metacone are aligned, with their bases adjoined. The paracone is present in all molars, while the metacone is only present in the M1-3 (it is completely absent in the M4, not even a vestigial cusp is present). The metacone is markedly larger than the paracone in the M3, but the size difference decreases anteriorly, so they are subequal in the M1. Both cusps are conical, circular in section in the M1, becoming subtriangular towards the last molar. The protocone is reduced and markedly lower than the para- and metacone, though more developed than in P. goini and T. atrox . Despite the difference in size, it is similar in shape to the protocone of P. goini , being anteroposteriorly compressed and lingually elongated. The state of preservation does not allow observation of the trigon basin, but it is apparently strongly reduced, similar to P. goini (with a vestigial trigon basin). The preservation condition does not allow the observation of either the paraconule and/or the metaconule.

The preparacrista is absent in the M1 and well developed, oblique to the labial edge in the M4. Its condition is obscured by wear in the M2-3. The portion corresponding to the centrocrista (postparacrista plus premetacrista) is strongly worn in the M1-3, in such a way that it is impossible to observe the crests.Due to the marked proximity between the paracone and metacone, these crests were likely very short.The postparacrista in the M4 is well-defined, short but more developed than in the precedent molars, and descends vertically through the posterior face of the paracone. The postmetacrista is strongly developed in the M1-3, long and oblique, similar to P. goini and less developed than in T. atrox . The wear degree does not allow determining the presence of a deep carnassial notch in this crest, as in P. goini and T. atrox .

The parastylar shelf is absent labial to the paracone in the M1 of A. gracilis , with the labial face of the tooth nearly vertical. It is present in the other molars, more developed in the M3 similar to P. goini ( Forasiepi & Carlini 2010) , but strongly reduced compared with non-thylacosmilid sparassodonts. In T. atrox , the parastylar shelf is absent in all molars ( Goin & Pascual 1987). On the M1, there is a tiny, blunt cusp located almost at the same height as the protocone, nearly aligned with the paracone and metacone (as in P. goini ), corresponding to the parastyle ( Forasiepi & Carlini 2010). As in P. goini , the parastyle on the M2 connects with a very short cingulum that descends towards the labial side of the tooth, forming a small ectocingulum (sensu Marshall 1978a), being better defined in the M3 ( Forasiepi & Carlini 2010). The metastylar shelf is present in the M1-3, increasing in size from the M1 to the M3, as in T. atrox and P. goini .

Lower dentition. There is no clear evidence of the number of lower incisors because the dentaries are broken at that portion in both the holotype and in specimen VPPLT-1612 ( Figs 2A ; 9 ; 11A, B ). The anterior margin of the left dentary of the specimen VPPLT-1612 is very cracked and poorly preserved ( Fig. 9A ). So, it is not clear if there was an alveolus is present immediately anterior to the canine. However, the possible space for incisors is strongly restricted, so they should be strongly reduced in size or set very tightly, and close to the canine, as in T. atrox . The presence of at least two lower incisors (in each hemimandible) is inferred for T. atrox , based on the position and size of those preserved (i.e., one at each side, with different position and size; Churcher 1985; Goin & Pascual 1987).

The lower canine is laterally compressed ( Fig.9A ; Appendix 5) but less than in T. atrox . Its orientation is oblique to the dental row in occlusal views, as in T. atrox ; and its implantation is sub-vertical (in lateral and lingual views; Figs 9A, B ; 11A ; char. 179), similar to other sparassodonts as Arctodictis , Australohyaena Forasiepi, Babot & Zimicz, 2015 , and Callistoe , but less than in T. atrox , which shows completely vertical implantation. The extra-alveolar portion of the lower canine shows a main, wide, and very shallow groove at its lingual face, with small lineal ridges on and near the main groove. Only the tip of the lower canine (c. ¼ of the extra-alveolar portion) is covered by enamel.

At least two lower premolars are identified, corresponding to p2-3, as interpreted for T. atrox ( Forasiepi 2009; Forasiepi & Carlini 2010; Suarez 2019). The dorsal portion of the dentary of the holotype of A. gracilis is broken anterior to the p2, and it is not possible to confirm the presence of an alveolus for the p1 ( Fig. 2A, B ). On the other hand, the left dentary of the specimen VPPLT-1612 shows a small opening (width = 1.58; length unknown) anterior to the p2, separated by a bone space (diastema?). It is roughly ellipsoidal, but its limits are broken, and the bone is poorly preserved, being unclear if it corresponds to an alveolus for a p1 or breakage. However, this small opening is not present in the right dentary, so we interpret that it is unlikely to be an alveolus. If that was the case, the p1 would be smaller than the p2, which is c. 49% shorter than the p3, showing a marked reduction. In the dentary of T. atrox , the p2-3 are similar in size and located closer to the molars than to the canine. In the specimen IGM 251108 from La Venta ( Goin 1997; Suarez 2019), there are three lower premolars: the p1 strongly reduced, almost vestigial, and the p2-3 much larger, being the p2 smaller than the p3. The p2-3 of A. gracilis are triangular in lateral view and asymmetrical, with the anterior edge of the cusp convex and shorter than the posterior ( Fig. 11E, F ), which is slightly concave. Both teeth are uninflated, with flat roots and placed obliquely to the main axis of the molars row, but aligned with the dentary, following its sigmoid geometry.

The lower molars show a marked posterior increase in size, common in sparassodonts ( Fig. 11 ). The posterior lobe of the crown is lower than the anterior one in m1-3 (seen in labial view), being strongly marked in the m2-4 (more in the m2) and less evident in the m1 ( Figs 2A, E ; 11C, E ). The main cusps in m1 are aligned in a single longitudinal row. The trigonid configuration in the m2-4 is open, with the paraconid in an antero-lingual position. The protoconid is the main cusp of the trigonid in all molars. The paraconid and protoconid increase in size posteriorly. A tiny structure is present in the m1, on the posterolingual face of the protoconid, which looks like a vestigial metaconid. However, this would differ from the two conditions seen in other borhyaeonoids: 1) the metaconid present only in the m2-4; or 2) absent (or vestigial) in all molars. However, a recently described specimen of Callistoe vincei ( Babot et al. 2022) shows another unusual condition, with a strongly reduced metaconid in the m3, probably also present in the m1 (inferred from a wear facet located on the mesiolingual corner of the trigonid); but “given the absence of a metaconid in the m2, that option is possible but unlikely” ( Babot et al. 2022: 475). The antero-lingual vertical crest of the paraconid (preparacristid interlocking mechanism) forms a keel. The precingulid is reduced, extended only on the base of the paraconid. There is a clear notch between these last two structures (hypoconulid notch) for contact with the preceding molar. The postparacristid and preprotocristid are well developed, at least in the m2-4, forming a carnassial notch between them. The postprotocristid is long, well defined and oriented to the antero-lingual corner of the talonid.

The trigonid is longer than the talonid in all molars (more than three to four times the length of the talonid), especially in the m2. The talonid is narrower than the trigonid in all molars and is reduced in relative size (compared to the trigonid) posteriorly, being almost vestigial in the m4, showing only a minuscule cuspid ( Fig. 11 ; Table 1 ). The talonid basin is slightly longer than wide in the m1, slightly wider than long in the m2 and markedly wider than long in the m3. The basin is divided into two portions: a labial one, with a concave and sub-horizontal surface, and a lingual one, markedly more vertical and flatter. The hypoconid is reduced in the m1-3, located approximately at the middle of the labial margin of the talonid. The entoconid is located in a posterolingual position. A vestigial hypoconulid is apparently present posterior to the entoconid and twinned with it. The bases of these two cuspids are merged, forming one block, although their tips are still differentiated. This lingual block of the talonid is laterally compressed and higher than the hypoconid, forming a verticalized lingual portion of the talonid basin. The preentocristid is well-developed and runs lingually to the trigonid. A labial postcingulid is present in the talonid of the m1-3, descending from the hypoconulid to the base of the labial face of the hypoconid. The overall talonid morphology (general morphology, basin morphology and cuspids distribution and morphology) is similar to that observed in the specimen from Quebrada Honda, referable to P. goini (see Material and methods, and Appendix 1, Material examined).

RESULTS

CLADISTIC ANALYSIS

The equal weights parsimony analysis produced six mostparsimonious trees, whose strict consensus is shown in Fig. 13 (Appendix 1, Figs A2 ; A 3 ). The consensus tree had a length of 1626 steps; CI = 0.306; and RI = 0.663; each of the most parsimonious tree with a length of 1617 steps, a consistency index (CI) of 0.307 and a retention index (RI) of 0.666. The implied weights analyses produced single trees for each case, obtaining the best score of 177.10957 with k = 3, 118.35478 with k = 6 and 71.89551 with k = 12 (Appendix 1, Figs A4 , A 5 ). The analysis using the k = 6 constant resulted in a similar topology to that where k = 3, with a few changes in the arrangement within Hathliacynidae Ameghino, 1894 and Lycopsis . The results of the Bayesian Inference analysis are presented in the Fig. 14 , showing the high posterior probability support values over 75%.

ECOMORPHOLOGICAL ANALYSIS

The body mass of 22.15 kg, obtained from the M2 area of the specimen VPPLT-1612, was selected as the best estimation because it presents the lowest value of %PE and best-adjusted R2 ( Table 2 ; see Material and methods). Additional estimations were made using other dentition variables with the next lowest %PE values, obtaining estimations between 22.29 kg and 27.29 kg ( Table 2 ). These estimations fall within the ‘large size’ category for South American carnivorous mammals ( Prevosti et al. 2013; see Material and methods).

Regarding to the diet inferences, a value of 0 was obtained for the RGA carnivory index, because the griding area is 0 (the talonid basin in the m4 is absent; see Material and methods), falling within the range of the hypercarnivorous mammals (i.e., those with diets including 70% or more of vertebrate flesh; see Van Valkenburgh & Koepfli 1993). The same value was obtained for T. atrox and P. goini . These results are congruent with those obtained for the relative length of the trigonid index calculated for the carnassial molar (m4) of A. gracilis (following Zimicz 2012; see Material and methods), between 0.92 (in the specimen VPPLT-1612) and 0.95 (in the holotype), falling within the category of meat-eater hypercarnivorous.

DIGITAL RECONSTRUCTION OF THE SKULL

The specimen VPPLT-1612 is deformed, with the right side of the skull shifted lower and the left side relatively raised due to taphonomic processes (more evident in anterior view; Fig. 15A, B). Although the right dentary has missing regions, the left is completely preserved, and there is no evidence of significant deformation in any of them (Fig. 15B). Based on the minor deformation in mandibles, it is inferred that the compression in anteroposterior and lateromedial directions of the skull was not strong. However, the compression degree in the dorsoventral direction could not be determined, so it was assumed that the dorsoventral deformation was not significant.

For the reconstruction presented here, the first step was a retrodeformation, raising the right side of the skull and lowering the left (see Material and methods), based on the less deformed hemimandible and particular anatomical structures (e.g., orbits, teeth row, external acoustic meatus; Fig. 15C, D). This retrodeformation did not pretend to restore with precision the original morphology of the skull, but instead the general shape of the skull with the aim to perform a life reconstruction of the head. After the retrodeformation, the mouth was opened to create a more informative appearance ( Fig. 16A ). The missing left upper canine and the right jaw were reconstructed from the corresponding structures on the opposite side (via mirroring; Fig. 16B ). The skull surface was smoothed ( Fig. 16B ) and the missing upper incisors and incomplete molars were reconstructed. Posteriorly, the masticatory muscles, eyeballs, and tongue were constructed on the skull ( Fig. 16B, C ). The masticatory muscles were based on modern opossums, and the eyeballs and tongue were built to fit in the orbits and the oral cavity. As there is no fossil evidence of the external appearance of A. gracilis (e.g., exterior details and fur color pattern), we used several mammalian taxa for reference: opossum, lion and leopard ( Fig. 16D ). In the first case, due to its closer phylogenetic affinities; in the remaining ones, because its ecological role falls closer to that of large modern felines (though not completely analogous; see discussion above). The reconstructed life appearance of the head of A. gracilis is shown in Fig. 17 .

Figures

FIG. 2. — Anachlysictis gracilis Goin,1997;IGM 184247 (holotype).Right mandibular ramus in lateral (A) and occlusal (B) views;C, left symphyseal flange fragment in medial and lateral views; D, left cranial fragment (postorbital portion; arrow indicates anterior direction); E, left mandibular ramus fragment with m2-3 in labial and lingual views; G, left m2-3 in occlusal view. Abbreviations: Fr, frontal; La, lacrimal; Na, nasal. Scale bars: 5 mm (vertical); 20 mm (horizontal).

FIG. 3. — Anachlysictis gracilis Goin, 1997; UCMP 39705, posterior fragment of right mandibular ramus, in lateral (A), posterior (B) and medial (C) views. Scale bar: 10 mm.

FIG. 4. — Anachlysictis gracilisGoin,1997;VPPLT 1612,skull in dorsal view (A) and lineal draw (B). Abbreviations:Ip, interparietal;Ju, jugal;La, lacrimal;Mx, maxilla; Na, nasal; Pa, parietal; Pal, palatine; pop, postorbital process; Px, premaxilla; smf, suprameatal foramen; Sq, squamosal. Scale bar: 20 mm.

FIG. 5. — Anachlysictis gracilis Goin, 1997; VPPLT 1612, skull in ventral view (A) and lineal draw (B). Abbreviations: Al, alisphenoid; Bo, basioccipital; Bs, basisphenoid; Eo, exoccipital; gf, glenoid fossa; hf, hypoglossal foramina; inf, incisive foramen;Ju, jugal; mpf, minor palatine foramen;mpp, medial palatine process of premaxilla; Mx, maxilla; n, notch; p, pits; pafo, paracanine fossa; Pal, palatine; pgp, postglenoid process; Psph, presphenoid; Pt, pterygoid; Px, premaxilla; Sq, squamosal. Scale bar: 20 mm.

FIG. 6. — Anachlysictis gracilis Goin, 1997; VPPLT 1612, left side of the skull: A, lateral view; B, dorsolateral view; C, lineal draw in dorsolateral view. Abbreviations: Al, alisphenoid; Eo, exoccipital; Fr, frontal; fro, foramen rotundum; iof, infraorbital foramen; Ip, interparietal; Ju, jugal; La, lacrimal;Mx, maxilla;Na, nasal; oc, occipital condyle; Os, orbitosphenoid; Pa, parietal; Pal, palatine; pgp, postglenoid process; pop, postorbital process; prgp, preglenoid process of jugal; sof, sphenorbital fissure; smf, suprameatal foramen; Sq, squamosal; scr, sagittal crest; tl, temporal line. Scale bar: 20 mm.

FIG. 7. — Anachlysictis gracilis Goin, 1997; VPPLT 1612, right side of the skull: A, lateral view; B, dorsolateral view; C, lineal draw in dorsolateral view. Abbreviations:Al, alisphenoid; dmm, depression for the masseter muscle; Eo, exoccipital; fdv, foramen for diploic vein; Fr, frontal; Ip, interparietal; Ju, jugal; La, lacrimal; Mx, maxilla; Na, nasal; oc, occipital condyle; Pa, parietal; Pal, palatine; pgp, postglenoid process; pop, postorbital process; prgp, preglenoid process of jugal; Px, premaxilla; smf, suprameatal foramen; Sq, squamosal; scr, sagittal crest. Scale bar: 20 mm.

FIG. 8. — Anachlysictis gracilis Goin, 1997; VPPLT 1612, posterior portion of the skull: A, detail of basicranium in ventral view; B, lineal draw of the basicranium in ventral view; C, occipital view of the skull. Abbreviations: Al, alisphenoid; Bo, basioccipital; Bs, basisphenoid; bjs, basijugular sulcus; cf, carotid foramen; Eo, exoccipital; fm, foramen magnum; fo, foramen ovale;gf, glenoid fossa; hf, hypoglossal foramina; ips, foramen for the inferior petrosal sinus;jf, jugular fossa; Ju, jugal; pap, paracondylar process of exoccipital; pgp, postglenoid process; prgp, preglenoid process of jugal; Pt, pterygoid; spht, sphenoidal tubercles (basilar tubercles); Sq, squamosal. Scale bar: 20 mm.

FIG. 9. — Anachlysictis gracilis Goin, 1997; VPPLT 1612, left dentary in lingual (A), labial (B) and occlusal (C) views; and right dentary in labial view (D). Abbreviations: acoc, anterior coronoid crest; an, angular process; con, mandibular condyle; cor, coronoid process; maf, masseteric fossa; manf, mandibular foramen; mc, masseteric crest (= inferior coronoid crest); mf, mental foramen; syf, symphyseal flange. Scale bars: 20 mm.

FIG. 10. — Anachlysictis gracilis Goin, 1997; VPPLT 1612, upper dentition. Right row in occlusal (A) and lingual (B) views; left side in occlusal (C) and lingual (D) views; and detail of the incisors area in occlusal view (E). Scale bars: 20 mm.

FIG. 11. — Anachlysictis gracilis Goin, 1997; VPPLT 1612,lower dentition.Left lower dentition in occlusal (A), labial (E) and lingual (F) views; and right lower dentition in occlusal (B), labial (C) and lingual (D) views. Scale bar: 20 mm.

FIG. 12. — Skulls of three species of Thylacosmilidae:A, Thylacosmilus atrox Riggs,1933;B, Patagosmilus goini Forasiepi & Carlini, 2010; C, Anachlysictis gracilis Goin, 1997. Scale bar: 20 mm.

FIG. 13. — Consensus tree obtained from parsimony cladistic analysis under equal weights. Bremer supports (from 6411 trees, cut 0) indicated over branches; Bootstrap support values indicated below branches. Supports not given for Dasyuromorphia, as this node was constrained a priori (see text).

FIG. 14. — Bayesian majority-rule consensus topology of the post-burnin sample of trees for selected taxa (eutherian outgroup and metatherian ingroup), including the thylacosmilid sparassodont Anachlysictis gracilis Goin, 1997, focus of this study. Posterior probability support values> 75% indicated in branches.

FIG. 16. — Sequential reconstruction of the skull of Anachlysictis gracilis Goin, 1997: A, modified skull with original mandible; B, smoothed skull after mirroring; C, skull with reconstructed eyes, masticatory muscles and a tongue; D, 3D model of a reconstructed head. Artist: Tatsuya Shinmura. Scale bar: 20 mm.

FIG. 17. — Reconstructed head of Anachlysictis gracilis Goin, 1997. Artist: Tatsuya Shinmura.

Tables

TABLE 1. — Dental measurements of the specimens of Anachlysictis gracilis Goin, 1997. Abbreviations: C/c, upper/lower canines; P/p, premolars; L, length; M/m, molars; Tal., talonid; Trg, trigonid; W, width. Measurements expressed in millimeters.

Specimen		C	P2	P3		M1	M2	M3	M4
VPPLT 1612 (right)	L	20.63	3.47	6.46		12.3	13.22	12.99	3.99
	W	6.74	1.88	4.42		6.97	8.83	10.2	10.8
VPPLT 1612 (left)	L	?	?	6.41		12.5	c. 12.60	c. 13.18	4
	W	?	c. 2.60	2.56		6.45	8.67	10.3	c. 10.70
		c	p2	p3		m1	m2	m3	m4
IGM 184247 (right)	L	?	c. 6.00	c. 7.20		c. 9.50	11.45	13.6	13.4
	W	?	c. 3.00	c. 3.30	Trg	c. 4.45	5.25	6	6
					Tal.	c. 4.00	4.75	5.05	1.75
IGM 184247 (left)	L	?	?	?		?	11.3	13.6	?
	W	?	?	?	Trg	?	5	5.8	?
					Tal.	?	4.45	5.1	?
VPPLT 1612 (right)	L	?	c. 4.25	c. 6.70		9.38	11.81	13.38	13.6
	W		c. 2.26	c. 3.91	Trg	4.25	5.6	6.59	6.73
					Tal.	3.6	4.85	5.41	2.5
VPPLT 1612 (left)	L	8.94	3.39	6.7		9.24	c. 12.05	c. 13.35	c. 14.00
	W	5.32	1.76	2.5	Trg	3.88	5.29	c. 6.46	6.23
					Tal.	c. 3.20	4.55	c. 5.62	2.5

TABLE 2. — Body mass estimations. Abbreviations: Log, common logarithm (with base 10); ln, natural logarithm (with base e); M2A, second upper molar area; M2L, second upper molar length; m2A, second lower molar area;m3L, third lower molar length; %PE, percent prediction error; R2, ratio estimate;SE, smearing estimate correction factor; X, selected variable. Body mass expressed in kilograms. Equation source references: 1, Myers 2001, ‘All species’ data-set; 2, Myers 2001, Dasyuromorphian data set; 3, Zimicz 2012.

						Body
Species	Equation	X	%PE	R2	SE	mass	Ref.
“ A. gracilis (IGM 184247)”	log(y) = 1.005 + 1.857 log(x)	m2A	7.00	0.951	1.119	22.77	1
	log(y) = 0.567 + 3.400 log(x)	m3L	12.00	0.945	1.035	27.29	2
	ln(y) = 1.76 + 3.17 ln(x)	m3L	12.08	0.950	1.110	25.29	3
“ A. gracilis (VPPLT 1612)”	log(y) = 0.426 + 1.890 log(x)	M2A	7.00	0.989	1.029	22.15	2
	log(y) = 1.005 + 1.857 log(x)	m2A	7.00	0.951	1.119	27.19	1
	ln(y) = 1.89 + 3.14 ln(x)	M2L	7.03	0.950	1.160	25.47	3
	log(y) = 0.567 + 3.400 log(x)	m3L	12.00	0.945	1.035	25.82	2
	ln(y) = 1.76 + 3.17 ln(x)	m3L	12.08	0.950	1.110	24.02	3
“ P. goini (MLP 07-VII-1-1)”	log(y) = 0.426 + 1.890 log(x)	M2A	7.00	0.989	1.029	25.93	2
	ln(y) = 1.89 + 3.14 ln(x)	M2L	7.03	0.950	1.160	19.79	3
“ T. atrox (P14531)”	log(y) = 0.426 + 1.890 log(x)	M2A	7.00	0.989	1.029	42.50	2
	ln(y) = 1.89 + 3.14 ln(x)	M2L	7.03	0.950	1.160	41.12	3