A M ERIC AN MUSEUM NOVITATES Number 3903, 42 pp. June 25, 2018 New Paleogene Notohippids and Leontiniids (Toxodontia; Notoungulata; Mammalia) from the Early Oligocene Tinguiririca Fauna of the Andean Main Range, Central Chile ANDRÉ R. WYSS,1 JOHN J. FLYNN,2 AND DARIN A. CROFT3 ABSTRACT Here we describe three new notohippid notoungulate species from the early Oligocene-aged Tinguiririca Fauna (Tinguirirican SALMA), recovered from volcaniclastic deposits of the Abanico Formation in the central Chilean Andes, two of which are known from material suffi- ciently complete to warrant formal naming. These include Eomorphippus bondi, sp. nov., a form of moderate size distinguished by hypsodont incisors and cheekteeth, as well as distinctive pro- portions of the upper incisors. A closely similar but more diminutive form is described as Eomorphippus neilopdykei, sp. nov. A third previously unrecognized notohippid in the Tinguirir- ica Fauna, best represented by a large, low-crowned, lower incisor battery, almost certainly rep- resents a new taxon, but remains too fragmentary to warrant naming now. We also propose a new binomial for a previously named notohippid, ?Eomorphippus pascuali, originally described from Gran Barranca in Argentina but which is now also recorded in Chile. This taxon, here named Rosendo pascuali, is markedly less hypsodont than E. bondi and E. neilopdykei and retains lingual cingula on at least p4–m1. As least one leontiniid notoungulate occurs in the Tinguiririca Fauna, Termastherium flacoensis, gen. et sp. nov., best represented by two partial upper cheek toothrows and a tentatively referred maxillary fragment bearing three deciduous teeth. Collec- 1 Department of Earth Science, University of California, Santa Barbara, and Division of Paleontology, Ameri- can Museum of Natural History. 2 Division of Paleontology, American Museum of Natural History. 3 Department of Anatomy, Case Western Reserve University School of Medicine, Cleveland, and Division of Paleontology, American Museum of Natural History. Copyright © American Museum of Natural History 2018 ISSN 0003-0082 2 AMERICAN MUSEUM NOVITATES NO. 3903 tively, description of these new fossils from Termas del Flaco, Chile helps to more fully character- ize the Tinguiririca Fauna, facilitating correlation and comparison to other South American land mammal faunas spanning the Eocene-Oligocene transition. INTRODUCTION Only in the past two decades has it been realized that the widespread volcaniclastic depos- its of the Andes of central Chile host one of the most important archives of Cenozoic mam- malian evolution in South America. The first mammalian fauna discovered in this region, the Tinguiririca Fauna of the upper Río Tinguiririca drainage (Novacek et al., 1989), has since been designated the type fauna of the Tinguirirican South American Land Mammal “Age” (SALMA) (Flynn et al., 2003). The Tinguirirican is currently regarded as temporally interposed between the Mustersan (late Eocene) and Deseadan (late Oligocene–early Miocene) intervals of the classical SALMA sequence, given that the Divisaderan, originally considered to precede the Tinguirirican, has recently been shown to be invalid as it represented a mixed assemblage of demonstrably older (mainly Casamayoran) and younger (Deseadan to Santacrucian) taxa (Cerdeño et al., 2008; López, 2008, 2010; López and Manassero, 2008). Following a preliminary accounting (Wyss et al., 1994), various components of the Tinguiririca Fauna have been described in detail elsewhere, including its marsupials (Flynn and Wyss, 1999), interatheriids (Hitz et al., 2000, 2006), archaeohyracids (Croft et al., 2003; Reguero et al., 2003), tardigrades (McKenna et al., 2006), dasypodids (Carlini et al., 2009), rodents (Bertrand et al., 2012), and notostylopids and basal toxodontians (Bradham et al., 2015). Here we build on this documenta- tion, describing the fauna’s diverse but sparsely represented notohippids and leontiniids. A spate of studies over the past two decades has enhanced understanding of the taxonomy and phylogeny of notohippid notoungulates. Noteworthy among these has been the description of one of the earliest known representatives of the clade, Pampahippus arenalesi from the Lumbrera Forma- tion of northwestern Argentina, and referral of several Casamayoran and Mustersan taxa from Patagonia, traditionally regarded as isotemnids, to the Notohippidae (Bond and López, 1993). Addi- tional species of Pampahippus have been described, P. secundus (Deraco and García-Lopez, 2016) and P. powelli (García-López et al, 2017). Shockey (1997) described three notohippids based on superbly preserved material from the Deseadan of Bolivia, providing the first cladistic parsimony analysis of relationships within the group. Gabbert (2004) described the basicranial anatomy of the Mustersan taxon, Puelia, contributing data relevant to unraveling higher-level toxodontian inter- relationships. A new late Oligocene notohippid from Quebrada Fiera (Mendoza, Argentina) was described in the context of a phylogenetic analysis of the more inclusive group to which it belongs (Cerdeño and Vera, 2010, 2014). López et al. (2010) documented notohippids from the “La Cancha” level (also referred to as a fauna and a locality) at the Gran Barranca in Patagonia, apparently Tin- guirirican in age, recognizing a large number of species (five) from that level. The names of the major subgroups of notoungulate mammals to which the species described below belong are not currently employed consistently. This nomenclatural instability has two primary sources. First, the continued application of traditional Linnaean ranks and differences 2018 WYSS ET AL.: NEW PALEOGENE NOTOHIPPIDS AND LEONTINIIDS 3 of opinion about which rank each group warrants have led to interchangeable endings of some names. Second, continuing uncertainty remains about which of the historically recognized notoungulate subgroupings constitute monophyletic entities, and thus warrant naming in the first place. Unfortunately, both of the species described below fall within one of the nomenclaturally most challenged branches of the notoungulate evolutionary tree. Of the four widely recognized notoungulate “subordinal,” groupings, the species described below are members of what is vari- ously termed the Toxodonta (Scott, 1905) or Toxodontia (Owen, 1853); see conflicting usages in, e.g., Simpson (1967), Mones (1986), and McKenna and Bell (1997). This group as a whole has yet to be diagnosed rigorously, and two of its five traditionally recognized subdivisions are demonstrably nonmonophyletic: Isotemnidae and Notohippidae (in contrast to Homalodotheri- idae, Leontiniidae, and Toxodontidae, each of which clearly are monophyletic). Although differing in some details, the analyses of Cifelli (1993), Shockey (1997), Cerdeño and Vera (2010), Billet (2011), and Shockey et al. (2012) have shown that the clade encompassing the species conventionally regarded as notohippids is nonexclusive, that is, the most recent com- mon ancestor (MRCA) shared by these species also gave rise to toxodontids and, potentially, leontiniids. Bond and López’s (1993) recognition of the notohippid affinities of various Casamay- oran and Mustersan species formerly regarded as isotemnids and the subsequent questioning of these results (Cerdeño and Vera, 2010) only underscore the need to stabilize toxodontian tax- onomy. The name Notohippidae, in particular, need not remain in such a confused state (referring simultaneously to differing sets of taxa). An obvious remedy would be to apply a phylogenetic definition, whereby some stem, node, or apomorphy would specify the clade to which the name is unambiguously linked. Formally proposing a phylogenetic definition for the name Notohip- pidae lies beyond the scope of the present analysis, however, given the poorly resolved under- standing of relationships among notoungulates in general and notohippids in particular. Nevertheless, we urge that Notohippidae ultimately be defined phylogenetically. In the interim, we employ the name Notohippidae informally, to refer to a monophyletic entity that approxi- mately encompasses the MRCA of Pampahippus and some particularly well-known, later-diverg- ing member of the clade such as Rhynchippus or Notohippus, plus all its descendants. Although with our current understanding of phylogenetic relationships this would make the Toxodontidae and Leontiniidae members of a clade bearing a name with an identical ending (-idae), such a circumstance is easily accommodated within a phylogenetic system. Alternatively, in a more traditional Linnaean classification, a higher-level clade name could be created for this group. Future analyses might also demonstrate homalodotheriids to be members of the Notohip- pidae (as the latter name is employed here), raising the question of whether the clade just described might more appropriately be named Toxodontia (in which case, the present report would more appropriately have been titled, “New Paleogene Toxodontians from…”). Neverthe- less, we have adopted the usage indicated above to avoid confusion with other uses of the name Notohippidae and to meet our immediate need to succinctly refer to the clade encompassing the species described below plus their nearest allies. We hope that this stopgap measure will be temporary, inasmuch as definitive phylogenetic definitions of this name and those associated with other clades of extinct South American ungulates are long overdue. 4 AMERICAN MUSEUM NOVITATES NO. 3903 ABBREVIATIONS Institutional abbreviations for specimens referred to in this study are: AMNH FM, Ameri- can Museum of Natural History, Division of Paleontology Fossil Mammal Collections; FMNH, Field Museum of Natural History; ICN-P, Instituto de Ciencias Naturales y Museo de Historia Natural, Universidad Nacional de Colombia, Bogotá; MLP, Museo de La Plata; and SGOPV, Museo Nacional de Historia Natural, Santiago. Dental abbreviations: L, left; R, right; i/I, lower/ upper incisor; c/C, lower/upper canine; p/P, lower/upper premolar; dp/P, deciduous lower/ upper premolar; and m/M, lower/upper molar. SYSTEMATICS Notoungulata Roth, 1903 Toxodontia Owen, 1853 Notohippidae Ameghino, 1894 Eomorphippus Ameghino, 1901 Type Species: Eomorphippus obscurus Ameghino, 1901. Diagnosis (emended from Simpson, 1967): Marked hypsodonty of the upper and lower incisors, posterior premolars, and molars. Cementum lacking. Upper incisors moderately pro- cumbent. I3 broad relative to I2 and I1. P4 more molariform than the anterior premolars, but lacking a distinct hypocone. Upper molars bearing hypocones, with a variable but deep cleft separating them from the protocone, the cleft blocked by the medial projection of what is likely the anterior end of the crochet. Other than the persistent major fossa, fossettes are obliterated early in wear. Simpson (1967) also listed P1–3 with notched or incomplete protolophs as diagnostic of Eomorphippus, but this feature applies to notohippids in general (Bond and López, 1993). He also considered lower molars with clefts anterior and posterior of the entoconid that variably develop into short-lived fossettids with wear as diagnostic of Eomorphippus, but this feature also appears to characterize a more inclusive group. Eomorphippus bondi, species novum Figure 1 Holotype: SGOPV 3046: partial skull preserving orbital rims, portions of both zygomatic arches, rostrum, and LI1–C, P3–M3; RI1–C, P3, M1–3. Paratype: SGOPV 2891, partial left lower dentition including i1–3 and p2–m3, plus rem- nants of the right dentition, including i1 and molds and external (labial) slivers of some pos- terior cheekteeth. Tentatively Referred Specimen: SGOPV 3085, labiolingually crushed right ?p4–m1. 2018 WYSS ET AL.: NEW PALEOGENE NOTOHIPPIDS AND LEONTINIIDS 5 Type Locality: The type, paratype, and tentatively referred specimen all derive from volcani- clastic sediments currently mapped as belonging to the Abanico (= Coya-Machalí) Formation in the Tinguiririca River valley (~35° S), in the Andean Main Range of central Chile, some 7 km west of the international boundary. (Geologic maps dating from before the late 1980s mistakenly identi- fied these deposits as belonging to the Cretaceous Colimapu Formation; Charrier et al., 1996.) All specimens described below were recovered from 35°–50° west-dipping strata, north of an unnamed 2738 m pass (indicated on the topographic sheet; Anonymous, 1985), approximately 3 km south of the town of Termas del Flaco, at what is termed the “main locality” in Charrier et al. (1996: fig. 6). Age: Early Oligocene (to potentially late Eocene), Tinguirirican SALMA. The diverse Tin- guiririca Fauna recovered from near Termas del Flaco, of which the notohippids are an impor- tant constituent, formed the basis for the formalized Tinguirirican SALMA (Flynn et al., 2003), which lies temporally between the Mustersan and Deseadan of the classical SALMA sequence. An extensive series of isotopic dates (summarized in Flynn et al., 2003) have been generated for strata hosting and underlying the Tinguiririca Fauna, indicating that E. bondi and its con- temporaries described below are no younger than ~31.5 million years old (early Oligocene) and could be ~1–2 million years older (Bradham et al., 2015). Etymology: In honor of Mariano Bond, for his enormous and influential contributions to the understanding of notoungulate phylogenetics and taxonomy. Diagnosis: Eomorphippus bondi generally resembles E. obscurus, differing from the latter mainly in being roughly 20% larger in most dental dimensions, in having a slightly smaller upper canine, and in having an I3 that is substantially wider than both I1 and I2. DESCRIPTION Eomorphippus bondi is the best-represented notohippid from the Tinguiririca Fauna and one of two relatively large-bodied species from the assemblage. The holotype has previously been referred to as Eomorphippus n. sp. (Wyss et al., 1994) and Eomorphippus undesc. sp., near E. obscurus (Flynn et al., 2003). Two specimens from Patagonia are of particular importance to our comparisons and identification of the new Chilean material. The first is MLP 12-1508, the holotype of Eurystomus stehlini Roth, 1901, collected by Santiago Roth from the enigmatic locality Cañadón Blanco (see below). This specimen consists of most of the upper and lower dentition and was referred by Simpson (1967), on Bryan Patterson’s recommendation, to Eomorphippus obscurus Ameghino, 1901. The second specimen is a nearly complete upper dentition that Egidio Feruglio collected from the Gran Barranca and deposited in the collec- tions of the University of Padua, Italy. Simpson (1967) referred this specimen to E. obscurus, a cast of which is deposited at the AMNH (AMNH FM 27885). Upper Dentition: A partial skull, SGOPV 3046, preserves a nearly complete upper denti- tion, serving as the sole record of the upper dentition of this taxon. This specimen has been slightly distorted through lateral compression, thereby slightly exaggerating the closeness of the posterior portions of the opposing toothrows, and possibly reducing the transverse dimen- sions of the cheekteeth, particularly the molars. Measurements are given in table 1. 6 AMERICAN MUSEUM NOVITATES NO. 3903 2018 WYSS ET AL.: NEW PALEOGENE NOTOHIPPIDS AND LEONTINIIDS 7 FIGURE 1. Photographs of cast, and line drawings of holotype of Eomorphippus bondi, SGOPV 3046, a partial skull bearing left I1–3, C, P3–M3 and right I1–3, P3, M1–3, in A, left lateral, B, C, occlusal, and D, anterior views (opposite page). Horizontal ridge in anterior view is a seam from the two-piece mold. Note hypsodonty of the incisors and narrowness of molars. Paratype of Eomorphippus bondi, SGOPV 2891, showing partial left lower dentition including i1–3 and p2–m3, in E, occlusal and F, labial views (above; photograph of cast). The three upper incisors form an impressive cropping apparatus arranged in a smooth arc. They are moderately procumbent, as is also well seen in AMNH FM 27885 (the Feruglio speci- men of E. obscurus; see Simpson, 1967: fig. 38). Lingual cingula (as have been noted for E. obscurus by Patterson in Simpson, 1967) appear to be absent, but this may simply reflect a fair degree of wear on I1-2 in SGOPV 3046. Rectangular wear facets occur on the lingual faces of the four central incisors. The centrally placed, blunt ridge present on the lingual face of I3 in E. obscurus (Patterson in Simpson, 1967) does not occur in E. bondi; instead, a broadly concave wear facet spans much of this portion of the tooth. All three incisors are moderately curved posteriorly along their substantial height. These teeth are faced anteriorly with broad, feature- less sheets of enamel. Breakage of the distalmost surficial layer of the premaxilla exposes much of the proximodistal height of these teeth, including large portions that were undoubtedly well within the alveolus (well above the gumline) in life. Significantly, enamel reaches nearly (if not completely) to the tips of the roots of these teeth (fig. 1 A, D). This remarkable degree of incisor hypsodonty is best seen on RI3, where the posterior margin of the tooth is exposed nearly to 8 AMERICAN MUSEUM NOVITATES NO. 3903 the tip of its root. The three incisors become progressively broader transversely, the mesiodistal diameter of I3 being roughly twice that of I1 (table 1, fig. 1D). The muzzle is broadest across the distal portions of the I3s, becoming progressively more constricted until roughly the ante- rior margin of P3, at which point the palate gradually widens again. The upper canine, preserved only on the left side, is greatly reduced in size in comparison to the incisors. Seemingly unworn, this tooth may not yet be fully erupted, as its tip is ~1 cm more dorsal than the bladelike termination of I3, although a similar relationship between the height of the canine and I3 is exhibited by AMNH FM 27885. The tooth is strongly canted anteriorly and situated posterior and slightly medial to the posterolateral margin of I3. The alveolus for a tooth of similar size and position is preserved on the specimen’s right side. The canine of SGOPV 3046 is just a sliver of a tooth compared to the incisors, much smaller than that of MLP 12-1508 and more similar to that of AMNH FM 27885—both of which Simpson (1967) referred to E. obscurus. Neither P1 nor P2 is preserved. A faint outline of an alveolus occurs midway in the gap between the canine and P3 on the left side of the specimen. Judging from the length of this gap and the size of the posterior premolars, no more than small spaces likely separated the teeth between the canine and the third premolar. A largely closed upper toothrow is also pres- ent in E. obscurus (Simpson, 1967). Both P3s are present and subtriangular in outline. It and all succeeding cheekteeth are moderately bowed. A strong paracone column or fold marks the high labial face of the tooth anteriorly. This structure, together with the small parastyle emanating from it, forms a strong anterolabial pillar. On the specimen’s left side, the posterior half of the labial face of this tooth is overlain by the parastyle of P4. This region of the tooth is exposed on the right side, since RP4 is missing, showing a subdued metacone fold. The ectoloph projects far beyond the rest of the occlusal surface of P3, which includes a large central fossa and a protocone forming its lingual border. Wear has obscured details of the loph arrangement, and evidence of any cingu- lum is lacking, in contrast with the condition in E. obscurus (Simpson, 1967). The P4 is more trapezoidal in outline than P3, with the tooth’s lingual side consisting of a broad, flat face. The tooth’s labial surface is basically a larger version of the arrangement seen on the preceding tooth (P3), except for its more prominent parastyle. The protoloph is com- plete and reaches the ectoloph anterolabially to form the flat lingual face of the tooth. The enamel near the lingual end of the posterior face extends slightly into the occlusal surface, and a minute fossette on the posterior crown surface suggests that a posterior cingulum may have been present earlier in wear. The anterior face of the tooth shows no evidence of a cingulum, again in contrast to the condition in E. obscurus (Simpson, 1967). The first two molars are quite similar in form, differing mainly in size and degree of antero- posterior elongation. All three molars are mildly imbricated. The paracone and metacone folds on the labial faces of the molars are less pronounced than in P4, but the molars all retain a short parastyle. Given the height of these teeth, their cross-sectional shape and dimensions would have changed substantially during wear. Beginning as mesiodistally elongate π-shaped structures, the exposed bases of the crowns indicate that a much squarer occlusal surface out- 2018 WYSS ET AL.: NEW PALEOGENE NOTOHIPPIDS AND LEONTINIIDS 9 line would have been achieved later in life. A strong cleft divides the protocone and the hypo- cone columns lingually. On M1, the protocone column makes up roughly two-thirds of the length of the concave lingual face of the tooth at the wear surface; on M2 the cleft is wider than on M1 at the level of the occlusal surface (narrowing dorsally toward the crown base). A small fossa on the hypocone column of M1 and M2 presumably partly represents the remnants of a posterior cingulum. M3 is incompletely erupted and little more than the anterior portion of its labial face and protocone column are visible. Lower Dentition: Most elements of the lower dentition are represented in SGOPV 2891 (fig. 1E, F). This specimen consists of a pair of toothrows, the left side of which is largely com- plete. The right side preserves only slivers and natural molds of the labial faces of the posterior cheekteeth and a poorly preserved i1. SGOPV 2891 is referred to E. bondi based on its compat- ibility in size and morphology to the upper dentition of the holotype. Of the incisor battery, Ri1 and Li1-3 are preserved but are somewhat damaged and not cleanly separated from the matrix. These teeth appear to be slightly dislocated ventrally with respect to the remainder of the toothrow, but since no trace of the mandibular ramus is pre- served, this is difficult to ascertain with confidence. All three incisors are spatulate in form, perhaps increasing slightly in mesiodistal breadth posteriorly. The lingual faces of these teeth remain covered in matrix, obscuring whether the conspicuous ridges seen in E. obscurus are present. The presence of wear facets is masked by poor preservation. The lower incisors are much more feebly developed than the uppers, the former measuring at most half the width of the latter. The lowers are arranged in a much tighter arc than their upper counterparts. Although the lower incisors are far less hypsodont than the uppers (the left i2, exposed clearly to the tip of its root, measures ~15 mm in height, while the right I2, which is incompletely exposed, measures >23 mm), enamel can be seen to cover not just the crowns, but also at least the labial faces of the roots. The lower incisor arcade exhibits none of the reduction in number or pattern of enlargement indicative of leontiniid affinities. Moreover, it is considerably narrower than the broad upper incisor/canine complex. This structural asymmetry between the upper and lower incisors is curious. In other respects, this lower dentition has the morphology and size expected for this species, and we have not recovered any lower dental elements in the Tinguiririca Fauna assemblage that would be better candidates for referral to E. bondi. A transversely oriented, crescentic sliver of enamel is present in a short, matrix-filled gap posterior to Li3; this probably represents a damaged remnant of p1, or a dislocated fragment of the canine. The remains of this tooth indicate that the diastema between the incisor series and the next most posterior tooth was no more than 5 mm long, and thus that the toothrow was largely closed. If a canine once was present, no clear evidence of it is preserved. The three posterior premolars, all little worn, increase in size posteriorly. A large, sharply angular trigonid dominates the crown of p2. The mesial end of the trigonid is slightly damaged, but a short, mesially projecting protolophid appears to have been present. There is no evidence of a paralophid. The metalophid is broad labiolingually, its apex form- ing a crest that is essentially transverse but mildly concave posteriorly. The lingual end of this crest is continuous posteriorly with a prominent cristid obliqua that delimits a feeble, 10 AMERICAN MUSEUM NOVITATES NO. 3903 TABLE 1. Mensural data for specimens described in text. Asterisk denotes uncertainty due to damage of the specimen in question. SGOPV i/I1 i/I2 i/I3 c/C p/P2 p/P3 p/P4 m/M1 m/M2 m/M3 Eomorphippus bondi, 3046 L Upper holotype AP 4.6 5.4 5.5 5.5 * 9.3 11.3 16.8 20.8 13.5* W 7.6 9.4 14.3 3.7* 7.2 8.8 11.2 9.6 6.5 R Upper AP 4.7 4.9 5.1 6.7 9.2 – 15 19.3 14 W 8.5 10.1 14.5 3.3 7.4 – 7.5 8.6 6.4 Eomorphippus bondi, 2891 L Lower paratype AP 5* 5* 12.5 14.5 20.4 23 27 W 6.3 6.3 8.9 9.8 8.5 Eomorphippus bondi, tent. 3085 R Lower ?p4 ?m1 AP 18.6* 21.7* W 7* 5* Eomorphippus neilopdykei, 2855 L Lower holotype AP 2.7 2.5 2.8 5.2 6.3 6.9 10.2 W 3.6 4.4 5.3 3.8 3.7 4.2 5.1 R Lower AP 2.5 2.3 2.6 5.2* 6.3* 7 10.2 12.6 15.2* W 3.6 3.7 5 3.7* 4.2* 4 5.1 5.1 6.9 * Rosendo pascuali, AMNH FM L Upper holotype 29405 AP 7.4 7.9 9.2 13.5 15.1 W 9.9 11 12.5 13.2 14.8 Rosendo pascuali AMNH FM R Lower 29474 AP 7.1 8.2 9.8 11.7 14.4 23.0 W 5.9 6 – 7.2 9.2 9.4 Rosendo pascuali 3051 L Lower AP 7.8 10.2 14.3 18.1* W 5.2 5.4 6.4 7* Rosendo pascuali 3096 L Lower AP 21.8 W 9.4 Rosendo pascuali 2991 R Upper AP 6.9 7.8 10.6 – – – W 7.7 8.4 11.6 – – –