The glyptodont Eleutherocercus solidus from the late Neogene of north-western Argentina: Morphology, chronology, and phylogeny

Glyptodonts (Mammalia, Xenarthra, Glyptodontidae) represent a diversified radiation of large armored herbivores, mainly related to open biomes in South America, with an extensive fossil history since the late Eocene (ca. 33 Ma) until their extinction in the latest Pleistocene–earliest Holocene. During the Pliocene and Pleistocene, glyptodonts arrived in Central and North America as part of the Great American Biotic Interchange. Within glyptodont diversity, one of the most enigmatic groups (and also one of the least known) are the Doedicurinae, mainly recognized by the enormous Pleistocene Doedicurus, with some specimens reaching ca. two tons. Almost nothing is known about the Neogene evolutionary history of this lineage. Some very complete specimens of the previously scarcely known Eleutherocercus solidus, which in turn becomes the most complete Neogene Doedicurinae, are here described in detail and compared to related taxa. The materials come from the Andalhuala and Corral Quemado formations (north-western Argentina), specifically from stratigraphic levels correlated to the Messinian–Piacenzian interval (latest Miocene–Pliocene). The comparative study and the cladistic analysis support the hypothesis that Doedicurinae forms a well supported monophyletic group, located within a large and diversified clade mostly restricted to southern South America. Within Doedicurinae, the genus Eleutherocercus (E. antiquus + E. solidus) is the sister group of the Pleistocene Doedicurus. Unlike most of the late Neogene and Pleistocene lineages of glyptodonts, doedicurins show along its evolutionary history a latitudinal retraction since the Pleistocene, ending with the giant Doedicurus restricted to the Pampean region of Argentina, southernmost Brazil, and southern Uruguay. This hypothetic relationship between body mass and latitudinal distribution suggests that climate could have played an active role in the evolution of the subfamily.


Introduction
Xenarthra is a peculiar clade of placental mammals characteristic for the Neotropical region (Gaudin and Croft 2015;Gibb et al. 2015;Delsuc et al. 2016), with a long fossil history since the early Eocene (Bergqvist et al. 2004;Lindsey et al. 2020). Their records are remarkably abundant in various Cenozoic sites mainly in South America (particularly in Argentina, see Scillato-Yané 1986), but also in Central America and North America (McDonald 2005;Brandoni et al. 2016;Gillette et al. 2016). Within Xenarthra, two large clades can be recognized, Cingulata and Pilosa, the latter containing the anteaters Vermilingua and the sloths Folivora (or Tardigrada or Phyllophaga, see Delsuc et al. 2001;Lindsey et al. 2020).
Within this diversity, glyptodonts (late Eocene-latest Pleistocene/earliest Holocene) are a clade composed of large to very large grazing armored herbivores, with body masses ranging between 100 kg to ca. 2000 kg (Vizcaíno et al. 2011;Soibelzon et al. 2012;Quiñones et al. 2020). In a phylogenetic framewok, there is consensus that pampatheres (Pampatheriidae) is the sister group of glyptodonts (Gaudin and Wible 2006;Gaudin and Lyon 2017;Fernicola et al. 2018; but see Delsuc et al. 2016;Mitchell et al. 2016 for an alternative view).
The early evolutionary history of glyptodonts, during the Paleogene, is poorly known, but the records increase markedly during the Neogene and Pleistocene (Gaudin and Croft 2015;Zurita et al. 2016), especially in southern South America (Zurita et al. 2016;Toriño and Perea 2018).
Though the diversity and phylogenetic relationships of glyptodonts are under study with promising results (see, among others, Fernicola 2008;Fernicola and Porpino 2012;Zurita et al. 2013;Gillette et al. 2016;Cuadrelli et al. 2019Cuadrelli et al. , 2020, the subfamily Doedicurinae (late Neogene-late Pleistocene) remains one of the most enigmatic groups. It is mostly known by the Pleistocene species Doedicurus clavicaudatus (Owen, 1847), one of the largest and most bizarre Quaternary forms, with some specimens having body masses of ca. 2000 kg (see Soibelzon et al. 2012). The most conspicuous characters of this clade include a caudal tube in which the distal part is dorso-ventrally compressed and laterally expanded, with terminal rugose concave areas, probably for the insertion of corneous "spines" (see Lydekker 1895: pl. 27). The fossil record (e.g., MLP 16-25) and recent biomechanical analyses show that this caudal tube would make a formidable weapon against predators or to be used in intraspecific combats (see Alexander et al. 1999;Blanco et al. 2009). Another intriguing character of Doedicurinae is the exposed surface of the osteoderms of the dorsal carapace, where large foramina cross the entire thickness of the osteoderms, a unique feature among glyptodonts, and even among Cingulates (see Zurita et al. 2014Zurita et al. , 2016, the most remarkable example being the late Pleistocene terminal species D. clavicaudatus. Finally, the geographic distribution along doedicurine evolutionary history is also peculiar, since a latitudinal retraction is especially evident in the Pleistocene genus Doedicurus Burmeister, 1874, with most records being restricted to southern South America (i.e., Argentina, Uruguay, and southernmost Brazil;Zurita et al. 2009Zurita et al. , 2014Varela et al. 2018).
Although the anatomy of Doedicurus is relatively well known, almost nothing is known about the Neogene diver sity achieved by the Doedicurinae in southern South America, with a fossil record mostly limited to one dorsal carapace, several fragments of associated osteoderms, and some caudal tubes (see Ameghino 1887Ameghino , 1889Ameghino , 1920Moreno 1888;Lydekker, 1895;Rovereto 1914;Castellanos 1927Castellanos , 1940Cabrera 1944). In a historical framework, and despite a large number of very poorly characterized species, only the species currently named Eleutherocercus antiquus (Ameghino, 1887) is known by a relatively complete dorsal carapace associated with a caudal tube (MLP 16-25, holotype of E. copei (Moreno, 1888) (see Ameghino 1887Ameghino , 1889Ameghino , 1920Moreno 1888;Lydekker 1895) originating from the early Pliocene of the Atlantic coast of Argentina (sensu Tomassini et al. 2013). In this context, Zurita et al. (2014) first described and included in a phylogenetic framework the two only known late Neogene Doedicurinae skulls associated with some osteoderms of the dorsal carapace coming from the late Neogene-earliest Pleistocene of the Pampean region of Argentina (ca. 4.5-2.8 Ma). The results indicate that the group including the genera Doedicurus and Eleutherocercus Koken, 1888 (Doedicurinae) is monophyletic, supported mainly by cranial and dorsal carapace synapomorphies (Zurita et al. 2014).
However, and despite the advance in the knowledge of these late Neogene Doedicurinae of the Pampean region of Argentina, other records of the subfamily from Northwestern Argentina, where one of the most complete late Neogene continental sequences is exposed, are very scarce (see Quiñones et al. 2019). In fact, the last revision of glyptodonts from this area was carried out by Cabrera (1944), and Doedicurinae were among the least addressed. According to this revision, only one species of Doedicurinae was recognized in the Neogene sequences, Eleutherocercus solidus (Rovereto, 1914), the type material (MACN 8335) being represented by osteoderms of the dorsal carapace, and coming from Santa María Valley (SMV).
During many years, the Santa María Valley was a valuable source of palaeontological data, and palaeontologists studied the abundant fossil remains discovered there (e.g., Moreno and Mercerat 1891;Ameghino 1891;Lydekker 1895;Rovereto 1914, among others in 1927in , 1929in , and 1930in (see Bonini 2014. During Riggs' field expedition, several remains of mammals including the most complete specimens of Glyptodontidae were obtained in the VQB, with precise data for their stratigraphic provenance. showing location of studied localities in Cata marca Province (A) and exposed lithostratigra phic units in the Villavil-Quillay Basin (B). Abbreviations: CQR, Corral Quemado River; HR, Hualfín River; JR, Jarillal River; LR, Loconte River; PCQ, Puerta de Corral Quemado; SFN, San Fernando Norte; VvR, Villavil River.
In this context, and as a result of fieldwork carried out in the upper Neogene from the Villavil-Quillay Basin, in addition to a careful revision of paleontological collections from the USA and Argentina, new and more complete specimens of Doedicurinae are described here, representing the most complete late Neogene glyptodonts ever known. This sample offers the opportunity to carry out a detailed morphological and phylogenetic study of these enigmatic cingulates, as well as to perform a comparative analysis with the relatively well known Doedicurinae from the Pampean region of Argentina, and to test their biostratigraphic importance.
The aims of this paper are: (i) to carry out a detailed description and comparison of new and more complete materials referred to E. solidus; (ii) to provide a taxonomic revision of this species and assess its biostratigraphic value; (iii) to test its relationships within Doedicurinae in a cladistic framework; and (iv) to discuss some aspects concerning the evolutionary history of Doedicurinae in high and middle latitudes of South America.

Material and methods
The analysis is mainly based on specimens FMNH-P 14437 and FMNH-P 14446, housed in the Paleontological collection of the Field Museum of Natural History, and MLP 29-X-10-29, housed in the Vertebrate Paleontological collection of Museo de La Plata.
The description of the units of the Santa María Group exposed in the Villavil-Quillay Basin (VQB; i.e., Puerta de Corral Quemado and San Fernando Norte localities) was based on different stratigraphical and geochronological pro-posals (Fig. 1B). Therefore, we perform here a theoretical model using this information, including stratigraphic, geochronological, and biostratigraphic data of the specimens of Doedicurinae found in VQB, in order to provide the chronological context of the fossils studied here. The provenance of the fossils collected by the expedition of Riggs in 1926 is provided in Marshall and Patterson (1981: appendix 5), while the correlation of the volcanic tuffaceous levels was made from the contributions of Latorre et al. (1997) and Bonini et al. (2017) (see Fig. 2).
In order to test the relationships of Eleutherocercus solidus within Doedicurinae, as well as the monophyly of Doedicurinae, we performed a cladistic analysis. The matrix includes 23 taxa and 57 morphological characters (see SOM, Supplementary Online Material available at http:// app.pan.pl/SOM/app66-Nunez_etal_SOM.pdf). Most characters are based on previous analyses (Fernicola 2008;Porpino et al. 2010;Zamorano and Brandoni 2013;Zurita et al. 2013Zurita et al. , 2014Zurita et al. , 2017Cuadrelli et al. 2020), with the addition of six new characters. The characters include five characters from teeth, thirteen from the skull, two from the mandible, three from the appendicular skeleton, two from the cephalic shield, twenty one from the dorsal carapace, and eleven from the caudal armor. The new characters are: angle between naso-frontal and parietal regions located at orbital notch level (in lateral view) (char. 5); angle between posterior margin of the orbital notch and the plane of the palate, near 90° (char. 7); small foramina with nearly homogenous distribution along the exposed surface of the osteoderms without ornamentation pattern (char. 34); posterior region of the dorsal carapace with a hypertrophied dome-shaped glandular structure (char. 38); transverse contour of the distal third of the caudal tube (char. 49); and ornamentation pattern of the dorsal osteoderms of the caudal tube (char. 51). There are 43 binary characters and 14 non-binary. All the characters considered in this study were scored via direct observation of the specimens and from photographs taken by the authors, and they were treated with the same weight (1.0). Character states that were not preserved were coded as "?". The matrix was obtained with Mesquite 3.04 (Maddison and Maddison 2018). The character-taxon matrix was analyzed via "Implicit enumeration" using TNT (Goloboff et al. 2008), under the criterion of maximum parsimony. Clade support was assessed via Relative and Absolute Bremer support (retained trees suboptimal by 4 steps; see Bremer 1994;Goloboff and Farris 2001); in addition to the Jackknife analysis, we used the option "Implicit enumeration" with 100 replicates.

Geological setting
The specimens described here come from late Neogene sediments cropping out in Puerta de Corral Quemado (PCQ) (27°13'S/66°55' W) and San Fernando Norte (SFN) (27°16' S/66°54' W) in the Villavil-Quillay Basin (VQB), Belén Department, Catamarca Province, north-western Argentina. These localities are included in the geological province known as north-western Pampean Ranges, and they are bounded to the NW by Sierra de Altohuasi and Culampajá, to the W by Cerro Durazno, to the SW by Sierra de Fiambalá, to the NE by Sierra de Hualfín, to the E by the Farallón Negro volcanic complex, and to the SE by the Cerro Pampa and Sierra de Belén (Fig. 1).
The glyptodonts studied here come from levels of the "Araucanense" and Corral Quemado horizons (sensu Riggs and Patterson 1939), cropping out in PCQ, which, with some differences at the boundary of units, were correlated with the Andalhuala and Corral Quemado formations, respectively (for more details about the correlations see Bossi et al. 1987;Bossi and Muruaga 2009;Bonini 2014;Esteban et al. 2014;Georgieff et al. 2017). Several authors reported radiometric datings obtained from the tuff levels interbedded in the Andalhuala Formation exposed in both localities (Marshall et al. 1979;Butler et al. 1984;Latorre et al. 1997;Sasso 1997;Bonini et al. 2017). These absolute ages enabled the correlation of both localities and constrained chronologically the fossiliferous levels, which highlighted their biostratigraphic value, as was indicated by Nuñez-Blasco et al. (2020). The specimens studied here and collected in SFN were exhumed from levels of the Andalhuala and Corral Quemado formations that span from ca. 4.8 to 3.6 Ma in this locality, according to the absolute ages proposed by Bonini et al. (2017). On the other hand, those coming from PCQ were found in levels ranging from ca. 5.64 to 3.66 Ma, following the absolute ages proposed by Latorre et al. 1997 (Fig. 2).

Historical background
The species currently known as Eleutherocercus solidus was first recognized and described by Rovereto (1914) as Neuryurus solidus. The holotype (MACN 8335) consists of a small fragment of dorsal carapace, exhumed from the Andalhuala locality, Catamarca Province. Some years later, Castellanos (1927) included the species in the genus Eleutherocercus Koken 1888, proposing a new combination, Eleutherocercus solidus (Rovereto, 1914). In the same contribution, Castellanos (1927) included another species in Eleutherocercus, namely E. tucumanus, originally Plaxhaplus tucumanus Castellanos, 1927, and designated as holotype the specimen MACN 2893, which consists of the distal part of a caudal tube, discovered in the vicinity of Tiopunco, Tucumán Province. Finally, Cabrera (1944) established Eleutherocercus tucumanus (Castellanos, 1927) as a junior synonym of Eleutherocercus solidus. Our observations are in agreement with the synonymy proposed by Cabrera (1944) since no significant differences are seen when comparing the caudal tubes of both species (see description and comparisons Emended diagnosis.-Medium sized glyptodont, and smaller compared to the giant Doedicurus clavicaudatus. Skull with the naso-frontal area ventrally inclined with respect to the parieto-occipital region, forming an angle of ca. 140° and showing some similitude with the genus Panochthus Burmeister, 1866. Orbital notch with vertical posterior margin, delimiting an angle with respect to the palatal plane of ca. 90°. Palate transversally expanded at the level of Mf1, but not as evident as in Doedicurus. Mf1 and mf1 simple and subcircular in outline, somewhat similar to those of Doedicurus; the remaining upper molariforms almost identical to Doedicurus. Dorsal carapace with the posterior region forming a "dome", perhaps of glandular origin. Exposed surface of the osteoderms of the dorsal carapace rugose and uniformly perforated by numerous and small foramina, a few of them crossing the entire thickness of the osteoderms; in Doedicurus instead, the foramina are larger and mainly concentrated in the central area of the osteoderms, with most of them crossing the entire thickness. Caudal tube similar to Doedicurus, but with its distal part not as laterally expanded; dorsal area of the caudal tube with numerous small foramina (as in the dorsal carapace), but preserving in some restricted areas a "rosette" ornamentation pattern (i.e., a central figure surrounded by several small figures).
Emended diagnosis.-Glyptodont similar in size to E. an tiquus (Table 1). Skull with an evident convex surface delimiting the rostral and parieto-occipital areas, more developed than in E. antiquus. Nasal openings sub-rectangular in contour, and somewhat different from the more curved lateral margins of E. antiquus and Doedicurus. In dorsal view, rostral area more elongated and subrectangular, different from the more subtriangular rostral area observed in E. antiquus and Doedicurus. Zygomatic arches with a great antero-posterior diameter. In lateral view, posterior margin of the orbital notch with an angle of ca. 90° to the palatal plane. In occlusal view, widening of the palate at the level of the Mf1 less developed compared to E. antiquus and Doedicurus. Very high horizontal ramus, almost identical to that of Doedicurus. Exposed surface of the osteoderms of the dorsal carapace very similar to E. antiquus and very different from Doedicurus. Caudal tube similar to E. antiquus in general morphology and preserving in some restricted areas a "rosette" ornamentation pattern, different from the completely perforated surface observed in Doedicurus.
Description.-Skull and dentition: The skull FMNH-P 14437 is almost complete and is the best preserved skull of Neogene doedicurine, since the two other coming from the upper Neogene of the Atlantic coast in the Pampean region do not preserve most of the parieto-occipital region (see Zurita et al. 2014) and since no cranial remains belonging to Doedicurinae are known outside Argentina.
In lateral view ( Fig. 3A 1 ), it can be observed that the dorsal profile of the skull resembles mainly that of cf. Eleutherocercus antiquus (MMP 4860) and, to a lesser extent, that of Doedicurus (MACN 2762, MMP 4251, MLP 16-24) (Zurita et al. 2014). The naso-frontal area is ventrally inclined to the parieto-occipital region, forming an angle of ca. 140°. This particular morphology is similar to that of the Panochthus, but in E. solidus this character appears behind the orbital notch, while in Panochthus it is just at the level  of the orbital notch. As in cf. E. antiquus, a conspicuous convex surface separates both regions of the skull, a potential synapomorphy for the genus. In turn, the orbital notch is slightly sub-elliptical, with the main axis dorso-ventrally oriented, but less developed compared to cf. E. antiquus and some specimens of Doedicurus (MMP 4251). In fact, the morphology of the orbital notch seems to be more similar between cf. E. antiquus and Doedicurus, compared to E. solidus. However, this morphology is different from that of Glyptodon and Glyptotherium, in which the orbital notch is more circular in outline. The ventral margin of the orbital notch shows a bony crest, very similar to that in cf. E. an tiquus and Doedicurus but less developed than in E. solidus.
In turn, the posterior margin of the orbital notch is remarkably vertical, drawing an angle with respect to the palatal plane of ca. 90°, as in cf. E. antiquus and Doedicurus. The descending process of the maxillae is more dorso-ventrally elongated than in cf. E. antiquus and much more elongated compared to Doedicurus, curving towards the posterior region of the skull, which is especially evident in the ventral third of this process. The transverse diameter of the zygomatic arch is constant from the region behind the orbital notch to the contact with the temporal bone, but the curved end is less developed than in Doedicurus. The zygomatic arch, remarkably antero-posteriorly developed, is very different from that of cf. E. antiquus and Doedicurus, in which it is clearly less antero-posteriorly developed but dorso-ventrally higher. The dorsal margin is straight and ventrally inclined towards the orbital notch, unlike cf. E. antiquus in which this dorsal margin delimits a more concave surface. In general, this results in a very different morphology for both species (i.e., E. solidus and E. antiquus).
In frontal view (Fig. 3A 2 ), the nasal openings are subrectangular, as in D. clavicaudatus (MLP 16-24), but different from cf. E. antiquus and Doedicurus sp. (MACN 2762), in which the lateral margins are somewhat convex, as in Plohophorus figuratus and Eosclerocalyptus tapinocepha lus (Cabrera, 1939). In general morphology, the outline of the nasal openings is very different from Glyptodon (G. re ticulatus and G. munizi), which shows a clear and inverted subtriangular outline (see Cuadrelli et al. 2019). In turn, the frontal bone shows a great vertical development, similar to cf. E. antiquus, and very different from D. clavicauda tus, Plohophorus figuratus, and G. munizi, in which it is clearly less developed. As observed in lateral view, the ventral margin of the orbital notch has a bony crest, similar in morphology to that of Doedicurus and much less developed compared to cf. E. antiquus. The infraorbitary foramina are sub-circular, and in the same relative position as in cf. E. antiquus and Doedicurus. In G. munizi and G. reticulatus, these foramina are circular, and developing a ventral channel (see Zurita et al. 2013;Cuadrelli et al. 2019) not present in Doedicurinae. The descending processes of the maxillae are in this frontal view very particular and different from those of cf. E. antiquus and Doedicurus. In E. solidus, this structure is much more massive, with a great transverse diameter (especially in its ventral half), also resembling D. clavicau datus. In Glyptodon, these processes have a similar transverse diameter, but in their ventral-most part, they become clearly pointed (see Cuadrelli et al. 2019Cuadrelli et al. , 2020. In dorsal view (Fig. 4A 1 ), the general morphology of the skull is similar to that of cf. E. antiquus. It is also somewhat similar to D. clavicaudatus. but some differences can be noted. In E. solidus, the rostral area ahead of the orbital notches is more antero-posteriorly elongated than in cf. E. antiquus and Doedicurus, delimiting a sub-rectangular contour. On the contrary, in D. clavicaudatus and Doedicurus sp., the rostral area is more sub-triangular, as in Panochthus tuberculatus and Plohophorus figuratus. As in cf. E. antiquus and Doedicurus, the orbital notches are posteriorly closed by a well-developed post-orbital bar. In E. sol idus, this post-orbital bar draws an angle of ca. 90° with the sagittal plane of the skull, while in the species of Doedicurus this angle ranges between 40-50°. The zygomatic arches of E. solidus are straight in their antero-posterior length, while in Doedicurus they are more laterally expanded, making a semicircle. The parietal and frontal bones are remarkably wide, especially between the orbital notches and the posterior-most region of the zygomatic arches. The maximum diameter of the fronto-parietal region coincides with the post-orbital bar. In cf. E. antiquus this morphology is similar, although this lateral expansion is not so evident compared to E. solidus. On the contrary, in other taxa such us D. clavi caudatus, Doedicurus sp., G. munizi, Panochthus tubercu latus, and Plohophorus figuratus, there is a well-developed post-orbital narrowing, weekly developed in E. solidus.
In occlusal view (Fig. 4A 3 , A 4 ), some of the most remarkable characters for E. solidus involve the morphology of the molariforms, the difference in size and shape between the Mf1 and the rest of the tooth series, together with the expansion of the palate at the level of Mf1. This set of characters is also seen in cf. E. antiquus, D. clavicaudatus, and Doedicurus sp., constituting potential synapomorphies for Doedicurinae (see Zurita et al. 2014). Mf1 are not lobed, as in Doedicurus and cf. E. antiquus. However, in E. solidus and Doedicurus it has a circular or sub-circular cross section, unlike cf. E. antiquus, in which Mf1 is more elliptical. Although there is a clear widening of the palate at the level of Mf1, this seems to be less obvious than in cf. E. antiquus and species of Doedicurus. The pattern of Mf2 tends to trilobation, while Mf3 is clearly trilobated. This feature is close to the morphology seen in Doedicurus, in which Mf2 already shows a high degree of lobulation, unlike in cf. E. antiquus where Mf2 is bilobated. The Mf3 of cf. E. antiquus and species of Doedicurus are similar, but in E. solidus and cf. E. antiquus the anterior margin of the first lobe is more rounded. In E. solidus, from Mf4 onwards, the trilobation is very marked and there are no significant differences between the other molariforms in the series, except that the anterior margin of the first lobe tends to be convex in Mf3-Mf4, flat in Mf5, and concave in the last three molariforms (Mf6-Mf8). In comparative terms, the Mf4-Mf6 of cf. E. antiquus, D. clavicaudatus, and Doedicurus sp. are very similar to each other, showing a concavity in the labial half of the posterior margin of the third lobe, a small detail also seen in other species of more distantly related genera, such as Plohophorus and Glyptodon. This character is also present in E. solidus, but beginning at the level of Mf5 instead of Mf4. In E. solidus, the minimum width of the palate is observed at the Mf3 and Mf4 level, while this narrowing is observed at the Mf4 and Mf5 level in cf. E. antiquus, and at the Mf4 level in Doedicurus. In cf. E. antiquus and Doedicurus, two large foramina with an anterior canal are present on the palate at the Mf3-Mf4 border, unlike in E. solidus in which those could not be detected. Likewise, in E. solidus the main axis of Mf1-Mf4 is parallel to the longitudinal axis of the tooth series, whereas in cf. E. antiquus only the main axis of Mf2 is parallel to the longitudinal axis, and in Doedicurus sp. only in the last molariforms of the series (Mf6-Mf8) are parallel. In other groups, such as Plohophorus and Glyptodon, this character is observed from Mf4-Mf8. It is worth noting that the skull MLP 29-X-10-9 presents supernumerary teeth, thus having a total of nine molariforms due to the duplication of Mf8. This feature was already described on this specimen by Cabrera (1944), making the first record of hyperdontia in Eleutherocercus and the second record for Glyptodontidae (see González-Ruiz et al. 2015).
In occipital view (Fig. 4A 2 ), the specimen FMNH-P 14437 is slightly deformed due to compaction, mainly affecting the maxillary bone. The dorsal outline of the supraoccipital is rounded and convex, similar to cf. E. antiquus, and a marked sagittal crest extends to the nuchal crest. In D. clavicaudatus and Doedicurus sp. the dorsal profile of the supraoccipital is slightly quadrangular and concave, and the sagittal crest cannot be seen in this view. In G. re ticulatus the dorsal profile of the supraoccipital is convex and continues laterally with the nuchal crest. In E. solidus, the foramen magnum is sub-elliptical, with its main axis transversally oriented. In D. clavicaudatus and Doedicurus sp. it is circular, while in G. reticulatus it is distinctly rhomboidal.  (Rovereto, 1914), FMNH-P 14437, from the late Miocene-early Pliocene of Argentina, Puerta de Corral Quemado, Andalhuala Formation, levels 23-28, Argentina; in dorsal (A 1 ) and occipital (A 2 ), and (A 3 ) occlusal views; drawing of detail of the first five molariforms (A 4 ). In white teeth the occlusal surface is covered by sediment.

Mandible:
The material includes FMNH-P 14437 (Fig. 5), represented by a complete and very well-preserved mandible, in addition to another hemi-mandible (FMNH-P 14446). The general morphology is almost identical to that of Doedicurus, although both taxa are temporally separated by more than 4 My. This highlights the conservative dentary morphology observed in this lineage, coinciding with the observations carried out in other glyptodont clades, such as the Neosclerocalyptini Neosclerocalyptus (see Quiñones et al. 2020).
In lateral view (Fig. 5A 1 ), the mandible is very robust, at the level of both the ascending and horizontal rami, as in Doedicurus. The ventral margin of the horizontal ramus is notably concave, much more than that of Glyptodon, Panochthus, and Neosclerocalyptus, reaching its maximum height at the level of the mf4 and mf5. The angle between the symphysial region and the middle and posterior border of the mandibular body is ca. 140°, as in Doedicurus, somewhat different from Glyptodon, in which this angle reaches 150°. Like the mandibular body, the ascending ramus is very robust, with a great antero-posterior diameter, which represents the total length of the tooth series from mf1 to mf6, as in Doedicurus. As in Doedicurus, Neosclerocalyptus, and Glyptodon, the ascending ramus is inclined forwards, drawing an angle of 65° between its anterior margin and the alveolar margin; this anterior margin is at the level of mf5, as in the genera Doedicurus and Glyptodon, and different from Neosclerocalyptus, in which it is located at the level of mf6. The sigmoid notch is very developed and similar to the notch in Doedicurus, while the posteroventral margin of the ascending ramus shows a very conspicuous angular process at the level of the posterior margin of the third lobe of mf8. In turn, the posterior margin of the ascending ramus is concave, especially in its dorsal-most third, as in Doedicurus, but different from Neosclerocalyptus, Panochthus, and Glyptodon, which show a straighter edge.
In occlusal view ( Fig. 5A 2 ), the pre-dental region of the symphysis is 76 mm long proximo-distally, widening in the most distal portion, where two large mental foramina are seen on the ventral side. The mandibular symphysis ends at the level of the second lobe of mf4, as in Neosclerocalyptus, Glyptodon, and Panochthus.
Molariforms: (Fig. 5A 2 ). The mf1 is sub-circular in section, and the smallest compared to the rest of the molariforms (mf2-mf8). In Doedicurus sp. and cf. E. antiquus this molariform is subcircular to subtriangular and similarly smaller than other molariforms. The mf2 of E. solidus is elliptical in outline, with irregular edges, lacking a defined pattern, so that no clear tendency towards bilobulation or trilobulation is observed. On the contrary, in Doedicurus sp., mf2 is completely trilobulated. This particular morphology observed in the mf1, mf2 of Doedicurinae markedly contrasts with that of Glyptodon, in which a very conspicuous trilobulation is seen from the mf1 (this character being a synapomorphy of the genus; see Zurita et al. 2013;Cuadrelli et al. 2020), together with a limited variation of size along the molariform series. The mf3 shows some tendency towards trilobulation, but not as marked as in the rest of the posterior molariforms. The mf4-mf8 are completely trilobulated, as in cf. E. antiquus; in addition, the general morphology of the last molariforms of E. antiquus and E. solidus is very similar to those of D. clavicaudatus and Doedicurus sp.
Osteoderms: Associated with the material described above, some fragments of dorsal carapace were found. This allows for careful comparisons with the holotype of E. sol idus (MACN 8335; see Fig. 6A) and, consequently, to refer the specimens here analyzed to this species, greatly improving its morphological characterization. Other remains were found in nearby localities, but mostly corresponding to isolated osteoderms and small fragments of carapace (MCH-P 188, 253, and 325; Fig. 6D). The fragment of dorsal carapace of FMNH-P 14446 (Fig. 6B) is composed of 11 associated osteoderms, of which six are complete. The osteoderms are hexagonal and probably correspond to the middle or cephalic part of the carapace, as observed in other glyptodonts. The exposed surface is rugose and has several small foramina irregularly distributed. Most of these foramina do not penetrate the entire thickness of the osteoderms, as in cf. E. antiquus (see Zurita et al. 2014) and the type material of E. solidus. This clearly differs from the morphology observed in Doedicurus, in which the osteoderms are much thicker, and the foramina tend to be, in small number, located in the middle region of the osteoderm, piercing its entire thickness. The other fragment (FMNH-P 14437; see Fig. 6C) belongs to the postero-dorsal region. The high degree of fusion between osteoderms prevents from inferring their exact number in this fragment. As in FMNH-P 14446 and the holotype of E. solidus, the exposed surface shows several small foramina irregularly distributed (see Rovereto 1914). The most marginal osteoderms are extremely irregular in shape and have an anomalous fibrous texture. On the left lateral half of this fragment, part of an irregular dome-like structure can be seen. The high degree of fusion indicates that this is the anterior region of the "dome". This particular structure (of glandular origin?) has been reported by other authors as characterizing the genus Eleutherocercus (see Moreno 1888;Lydekker 1895;Ameghino 1920;Chimento et al. 2010). We therefore suggest, agreeing with the proposal of Chimento et al. (2010), that it could represent part of a glandular system located at the level of the posterior part of the pelvic girdle. Supporting this hypothesis, several specimens of the Glyptodontinae Glyptodon show very large foramina at the intersection between the annular and radial sulci, in the same region of the dorsal carapace (see Zurita et al. 2016;Cuadrelli et al. 2019).
pattern. However, its general morphology can be inferred. Although its transverse diameter is quite similar along its antero-posterior length, the caudal tube is slightly more laterally expanded at the proximal end, whereas the distal-most part is clearly rounded and blunt. In lateral view, the caudal tube shows a greater dorso-ventral diameter (ca. 30%) in its proximal part, being more flattened towards the distal end. MLP 29-X-10-21 (Fig. 7A) and MACN 2893 (holotype of Fig. 7. Caudal tubes of doedicurine glyptodont Eleutherocercus solidus (Rovereto, 1914), from upper Miocene-Pliocene of Argentina. A. MLP 29-X-10-21, from Puerta de Corral Quemado (near the road to Loconte), unknown stratigraphic level; in lateral (A 1 ) and dorsal (A 2 ) views; drawing of detail of dorsal osteoderm (A 3 ). B. MACN 2893 (holotype of E. tucumanus Castellanos) from Tiopunco, unknown stratigraphic level; in lateral (B 1 ) and dorsal (B 2 ) views; drawing of detail of dorsal osteoderm (B 3 ). Fig. 7B) preserve the exposed surface, with a clear ornamentation pattern. In dorsal view, the ornamentation displays a "rosette" pattern. In each osteoderm the central figure is slightly concave, surrounded by 7-8 smaller and angular peripheral figures. The principal sulcus surrounding the central figure bears several small foramina. In some specimens (e.g., MLP 29-X-10-21) the foramina are more developed in the proximal half of the caudal tube, while in others (e.g., MACN 2893) they are present along the antero-posterior diameter, but this could be probably due to taphonomic factors. On the lateral sides, near both margins, there is a row of large figures with a very rugose surface, which become more numerous towards the distal end of the tube. In lateral view, six or seven lateral figures are present, increasing in size towards the distal end, as in Panochthus and Doedicurus. Unlike in Panochthus, in Eleutherocercus (as well as in Doedicurus) the surface of these figures is concave, very rugose and striated. Instead, in Panochthus and Hoplophorus these figures display a "spine-like" morphology. In E. solidus, lateral figures are separated from each other by a single row of foramina, but the proximal ones also show some peripheral figures. On the distal end of MACN 2893, there are two large figures on each margin, one being more dorsal and the other being more ventral and somewhat larger. The contact area between both figures (dorsal and ventral) coincides with the midline of the lateral sides of the tube. Despite some similarities with Doedicurus, the morphology of the caudal tube of Eleutherocercus shows some differences. In Doedicurus, the distal end of the caudal tube is very laterally expanded, similar to a club. The rosette ornamentation pattern of Eleutherocercus is not observed in Doedicurus; in the latter, the surface is smooth with numerous foramina similar in morphology to those of the dorsal carapace.

E. tucumanus,
However, the most evident differences between the genera Doedicurus and Eleutherocercus are observed at the distal end of the caudal tube, specifically in the location and number of the large rugose figures. In Eleutherocercus these figures are numerous, ranging 10-14 depending on the specimen. From the most distal end of the tube to a proximal position they are located as follows: four at the most distal end, grouped into two pairs, one pair placed in the dorsal surface, and the other two in the ventral surface, the latter being a larger pair; in the dorsal area of the distal end there is a total of up to five figures, grouped into a distal row of three and another more proximal composed of two figures. In turn, the ventral area shows the same ornamentation pattern to that observed in the dorsal surface. On the other hand, in Doedicururs these figures are reduced in number to only 8. From the most distal end of the tube to a proximal position they are located as follows: a pair of symmetrical figures at the most distal end of the caudal tube; four immediately behind, in a more proximal position and grouped in two pairs, one pair in a dorsal position and the other in a ventral position; and finally two more symmetrical figures in a mid-lateral position, the latter being the largest.

Discussion
Glyptodonts constitute one of the most spectacular cingulate radiations in America (Zurita et al. 2016), and their evolutionary history appears to be mostly related to the appearance and development of open biomes since the late Eocene-Oligocene Mitchell et al. 2016;Zachos et al. 2001Zachos et al. , 2008Fig. 8). The biomechanical evidence interpreted glyptodonts as grazers, although some kind of niche partitioning was suggested by some authors .
Although the Paleogene evidence is still very scarce and mostly limited to osteoderms of the dorsal carapace, an interesting diversification is observed, represented by two morphotypes, one included in the subfamily "Propalaehoplophorinae" and the other in the poorly known "Glyptatelinae". This early diversification could be related to the progressive aridization and development of open areas since the latest Eocene-early Oligocene, and is mostly restricted to the Patagonian region of South America .
Since the early-middle Miocene (ca. 19-17 Ma) a marked increase in the records of glyptodonts is observed, with the first well-documented Neogene radiation represented by the "Propalaeoplophorinae" and the enigmatic Parapropalaehoplohorus septentrionalis in southern South America (Croft et al. 2007;Vizcaíno et al. 2010;González Ruiz et al. 2020). This evolutionary process may have coincided, at least in part, with the Middle Miocene Climatic Optimum (MMCO) (Zachos et al. 2001;Croft et al. 2016).
From an evolutionary standpoint, the "Propalaehoplophorinae" (i.e., Propalaehoplophorus, Eucinepeltus, and Cochlops) are the most basal taxa of a large southern South American radiation that includes most of the known diversity of glyptodonts, the other large radiation being represented by the Glyptodontinae (Carlini et al. 2008;Zurita et al. 2013;Cuadrelli et al. 2020). In this framework, this first large southern lineage that begins with Propalaehoplophorus australis, underwent, according to the fossil record, a diversification process since the late Miocene-earliest Pliocene, with the first record of several lineages, some of them with long biochrons that reached the latest Pleistocene (Zurita et al. 2016). Interestingly, this major diversification seems to have mainly occurred in high latitudes instead of tropical and intertropical regions of South America, where the diversification was mostly limited to the subfamily Glyptodontinae during the Miocene and Pliocene (Zurita et al. 2013;Cuadrelli et al. 2020).
In this context, in our phylogenetic analysis, Doedicurinae appears as a well supported clade, including until now the genera Eleutherocercus and Doedicurus as sister groups.
In turn, Doedicurinae is included in a large southern South American clade, which is in agreement with several previous analyses (see Zurita et al. 2013;Cuadrelli et al. 2020; but see Fernicola 2008; Fernicola and Porpino 2012 for another interpretation). As occurs with most of the taxa composing this large clade, the evolutionary history of Doedicurinae was restricted to high and middle latitudes.
From a chronostratigraphic view point, the records of E. solidus come from the Corral Quemado and Andalhuala formations, specifically, the stratigraphic levels chronologically located in a latest Messinian to Piacenzian interval (ca. 6-3 Ma, latest Miocene-Pliocene). FMNH-P 14475 is considered here as the First Appearance Datum (FAD) of Eleutherocercus solidus in VQB, recorded in level 20 (sensu Stahlecker in Marshall and Patterson 1981: appendix 5), which is located in the stratigraphic section 1100-1200 m at ca. 6 Ma (see Fig. 2). This inferred interval is similar to the one postulated for the geological units in the Pampean region of Argentina where remains of E. antiquus (Monte Hermoso Formation) and cf. E. antiquus (Chapadmalal and El Polvorin formations) are recorded, which span ca. 5.2-2.8 Ma (Zanclean to Piacenzian) (Zárate et al. 2007;Beilinson et al. 2017;Quiñones et al. 2020). This implies that both species of the genus partially overlap in time (Pliocene), but in different regions of southern South America (Zurita 2007;Cione et al. 2000).
During the late Miocene-early Pliocene, a series of paleoenvironmental changes have been documented in the fossiliferous sediments of northwestern Argentina. These changes have been related to floristic (i.e., C3/C4 ratio), tectonic (i.e., rising of mountain ranges), and climatic events (i.e., generalized aridification, seasonality) (Hynek et al. 2012;Georgieff et al. 2017;Armella and Bonini 2020), which probably affected the development and evolution of herbivorous faunas (Cerling et al. 1998;MacFadden et al. 1999;MacFadden 2005). The FAD of E. solidus in VQB is synchronous to large eolian deposits representing arid conditions, which are identified in different locations along the basin (see Muruaga 2001;Hynek et al. 2012;Esteban et al. 2014;Georgieff et al. 2017). Likewise, these deposits coincide with the period in which C4 becomes an important component in the diet of large herbivores (Hynek 2011), before becoming dominant after 4 Ma (Latorre et al. 1997). As stated above, the general morphology of the mandible and teeth of E. solidus is almost identical to that of Doedicurus, which has the highest hypsodonty index among glyptodonts (Vizcaíno et al. 2011). These anatomical features can be interpreted as a way of increasing the mechanical capacity in relation to accidental consumption of abrasive particles (i.e., sand, dust, volcanic glass) adhering to the surface of plants (Janis 1988;Candela and Bonini 2017) or the ingestion of large amounts of food with low nutritive value (Vizcaíno et al. 2011;see Fig. 9).
The high frequency of records suggests that Doedicurinae were dominant elements in the late Neogene palaeofauna in both the Pampean and northwestern regions. However, preliminary studies of the El Polvorín Formation (Pliocene-earliest Pleistocene) reveal a decrease in the frequency of records of Doedicurinae towards the Pleistocene (Quiñones et al. 2020). In this framework, and taking into account that most of the glyptodonts present a rich fossil record in South America (see Zurita et al. 2009), we suggest that this very low frequency of records of Pleistocene Doedicurinae in most of South America (perhaps except for some areas of Uruguay) is reflecting a genuine process in which Pleistocene Doedicurinae were very reduced in the number of specimens compared to the other genera (ie., Glyptodon, Panochthus, and Neosclerocalyptus). It is interesting to note that it is quite common to find some associated specimens of Pleistocene glyptodonts in southern South America, suggesting some kind of gregarious habit. However, this was not reported for Doedicurus since all the records correspond to single specimens. Although more evidence is necessary, this could suggest that Doedicurus developed solitary habits.
Outside Argentina, Neogene Doedicurinae is also recorded in the Pliocene of Inchasi, Bolivia (ca. 3.3-4 Ma; see MacFadden et al. 1993;Cione and Tonni 1996), and Uruguay (Koken 1888;Toriño and Perea 2008), at high and middle latitudes. Doedicurinae (together with Glyptodontinae) were some of the most frequently recorded glyptodonts in the Pliocene sequences of Inchasi, Bolivia (AEZ, personal observations). The northernmost record of a Doedicurinae comes from the Pliocene of Ayo Ayo, near La Paz (Bolivia), at a latitude of ca. 16°30´. Based on the high frequency of Neogene records of Doedicurinae in Uruguay, some authors recognized several endemic taxa, among them Prodaedicurus and Castellanosia (Toriño and Perea 2008). However, it cannot be ruled out that some (if not all) could belong to the same taxon. More complete remains are needed to test this taxonomic hypothesis. Indeed, unlike north-western and Pampean regions of Argentina, where Doedicurinae is repre-sented by several relatively complete specimens, the remains from Bolivia and Uruguay are mostly represented by osteoderms of the dorsal carapace and some caudal tubes, a situation that hinders their taxonomic identification. No remains of Doedicurinae are known from the Pleistocene of Bolivia and Peru. However, it seems that during the late Neogene (latest Miocene-Pliocene) Doedicurinae reached a wide latitudinal distribution from ca. 16°30´ (La Paz, Bolivia) to ca. 38°58´ (Monte Hermoso, Argentina), being a frequent component within glyptodont assemblages.
It is worth noting that the largest (Doedicurus) and the smallest (Neosclerocalyptus) glyptodonts were the most latitudinally restricted during the Pleistocene, with Doedicurus being the most geographically constrained (Zurita et al. 2009;Quiñones et al. 2020). This progressive latitudinal retraction of Doedicurinae since the Pleistocene is quite intriguing because it is exactly opposite to that of the other Pleistocene lineages (e.g., Glyptodon and Panochthus) of South American glyptodonts. In the latter, an increase in body mass is related to an increase in the latitudinal distribution (see Cuarelli et al. 2020), as is currently observed in some African megamammals (i.e., Hippopotamus am phibius) that have a body mass similar to that inferred for Doedicurus (ca. two tons; Soibelzon et al. 2012;Vizcaíno et al. 2011). The co-ocurrence of Doedicurus and at least three other genera of glyptodonts (i.e., Neosclerocalyptus, Fig. 9. Hypothetical life reconstruction of the extinct Eleutherocercus solidus (Rovereto, 1914), based on the new specimens described here. Credit: Pedro Cuaranta.
Panochthus, and Glyptodon) in the Pampean region of Argentina, Uruguay, and southern Brazil, suggests niche partitioning, in agreement with evidence from stable isotope analyses (see Domingo et al. 2012;Melo-Franca et al. 2015) and biomechanical studies (see Vizcaíno et al. 2011;Pomi 2008). According to Cione et al. (2015), the hypothetically low primary productivity of these glacial ecosystems could have led to a lower number of individuals for each species, also stimulating a partition of resources as a key strategy to survive in these areas, especially during glacial stages (see Domingo et al. 2020).
Accordingly, our observations reveal that the frequency of records of Doedicurus is one of the lowest within Pleistocene glyptodont diversity in southern South America. Therefore, it is possible that the number of individuals of Doedicurus living in the same area at the same time was low, as suggested by the fossil record.
This geographic retraction, which probably started in the latest Neogene or early Pleistocene, could be related to the progressive cooling and aridization observed since the Pliocene, which intensified during the Pleistocene in southern South America. Despite this phenomenon, Doedicurus is one of the last survivors of the South American Pleistocene megafauna, as it is recorded until the latest Pleistocene in the Pampean region, a pattern that suggests that this area acted as a refuge for the last member of the megafauna before its extinction (see Politis et al. 2019).

Conclusions
(i) Eleutherocercus solidus is the only well characterized Doedicurinae in North-western Argentina, with a biochron that spans from the latest Miocene to Pliocene; (ii) the mandibular and dental morphology of E. solidus (almost identical to Doedicurus, the most hypsodont glyptodont ever known) reveals that the temporal distribution of this species coincides with a spread of C4 plants in northwestern of Argentina; (iii) the other species of the genus, E. antiquus, shows a similar temporal distribution (Pliocene), but living in a different biogeographical area (i.e., the Pampean region of Argentina); (iv) the Eleutherocercus is a well supported clade, sister group of the giant late Pleistocene Doedicurus clavicaudatus; (v) Eleutherocercus spp. + Doedicurus constitutes the subfamily Doedicurinae, which in turn forms part of a large southern South American glyptodont radiation starting, at least, in the early Miocene, in the Patagonian region of Argentina; (vi) contrary to what was observed for the diversity of the other Pleistocene glyptodonts (i.e., Neosclerocalyptus, Panochthus, and Glyptodon), Doedicurinae underwent a latitudinal retraction ending with the giant Doedicurus being limited to the Pampean region of Argentina, Uruguay, and southernmost Brazil during the late Pleistocene; (vii) the hypothetic relationship between body mass (with some specimens of Doedicurus achieving ca. two tons) and latitudinal retraction suggests that the cyclic climate changes during the Pleistocene could have played an active role in the final steps of the evolutionary history of glyptodonts in South America.