Description of a fossil camelid from the Pleistocene of Argentina, and a cladistic analysis of the Camelinae

We describe a well-preserved South American Lamini partial skeleton (PIMUZ A/V 4165) from the Ensenadan (~ 1.95–1.77 to 0.4 Mya) of Argentina. The specimen is comprised of a nearly complete skull and mandible with full tooth rows, multiple elements of anterior and posterior limbs, and a scapula. We tested this specimen’s phylogenetic position and hypothesized it to be more closely related to Lama guanicoe and Vicugna vicugna than to Hemiauchenia paradoxa. We formulate a hypothesis for the placement of PIMUZ A/V 4165 within Camelinae in a cladistic analysis based on craniomandibular and dental characters and propose that future systematic studies consider this specimen as representing a new species. For the first time in a morphological phylogeny, we code terminal taxa at the species level for the following genera: Camelops, Aepycamelus, Pleiolama, Procamelus, and Alforjas. Our results indicate a divergence between Lamini and Camelini predating the Barstovian (16 Mya). Camelops appears as monophyletic within the Camelini. Alforjas taylori falls out as a basal member of Camelinae—neither as a Lamini nor Camelini. Pleiolama is polyphyletic, with Pleiolama vera as a basal Lamini and Pleiolama mckennai in a more nested position within the Lamini. Aepycamelus and Procamelus are respectively polyphyletic and paraphyletic. Together, they are part of a group of North American Lamini from the Miocene epoch. Electronic supplementary material The online version of this article (10.1186/s13358-020-00208-6) contains supplementary material, which is available to authorized users.


Introduction
Exceptional fossils are key to solving taxonomic and phylogenetic questions given the rich and reliable anatomy they preserve. Some of them are still waiting in the field, while some are in museum collections in which proper curation and care allow multiple researchers to re-assess them. We present a case of the latter, with the study of specimen PIMUZ A/V 4165 from Barranca del Parana, San Nicolas (Buenos Aires province, Argentina) (Roth 1889;Schulthess 1920), part of a collection assembled by the celebrated Swiss-Argentinian paleontologist Santiago Roth, 1850-1924(Bond 1999. between the late Hemingfordian and early Barstovian , during the fourth radiation. Honey et al. (1998) estimated a divergence between the Lamini and Camelini during the . However, the timing of this split is still ambiguous, partly due to systematically problematic taxa, including the genus Aepycamelus. Honey et al. (1998) attributed Aepycamelus (late Hemingfordian to late Hemphillian, ~ 17. [5][6] to the Lamini, while observing that members of this genus could also be ancestral to both Lamini and Camelini. Adding to the discussion, studies based on mitochondrial data estimated the divergence of Lamini and Camelini to approximately 25 Mya, during the Arikareean (Cui et al. 2007).
The generic diversity of camelids decreased during the late Miocene when Aepycamelus and the last members of the Protolabinae are last recorded ). The last North American camelids disappeared in the late Pleistocene (Kurtén and Anderson 1980) (Fig. 1).

Previous studies on camelids
One of the first morphology-based phylogenies including a broad range of camelids was that of Webb (1965), which also provided an osteological description of Camelops. Honey and Taylor (1978) later revised the relationships among Protolabinae. In 1979, Harrison described the genus Alforjas and suggested a phylogeny for the Camelinae, later refining it with the inclusion of North American extinct giant Camelini (Harrison 1979. A few years later, Honey et al. (1998) proposed an extensive taxonomic revision of the entire Camelidae clade. They attributed the following extinct genera: Aepycamelus, Blancocamelus, Hemiauchenia, Palaeolama, Alforjas, Camelops, and Pliauchenia, to the Lamini and Procamelus, Megatylopus, Titanotylopus, Megacamelus and Gigantocamelus to the Camelini. Webb and Meachen (2004) later invalidated the genus Pliauchenia and re-positioned some of its members within a new genus, Pleiolama, along with two newlydescribed species: Pleiolama mckennai and Pleiolama vera. In the same work, they also described Alforjas magnifontis. Scherer (2013) proposed the most recent phylogeny based on a cladistic analysis of a broad range of South American species and new postcranial characters. She considered the following South American taxa as valid: Hemiauchenia paradoxa, Palaeolama major, Fig. 1 Time range of extinct species included in this study. Not to scale, species time range approximate. Species positioning in the Protolabinae follows  and Honey (2007). Species positioning in the Camelinae, Lamini, and Camelini follow our results. Species range for North American species following Baskin and Thomas (2016); Bravo-Cuevas and Jiménez-Hidalgo (2015); Harrison (1979); Honey and al. (1998);Honey and Taylor (1978); Pagnac (2005); Peterson (1911); Prothero (2005); Webb and Meachen (2004). Absolute ages for North American land mammal age following Woodburne (2004). Species range for South American species are uncertain and discussed by Scherer (2013). Absolute ages for South American land mammal age following Verzi et al. (2004); Soibelzon and al. (2008); Cione and Tonni (2001); Cione and Tonni (2005); Cione et al. (2015). Abbreviations: Chadron. = Chadronian; Whitn. = Whitneyan; Arikar. = Arikareean; Heming. = Hemingfordian; Barstov. = Barstovian; Clarend. = Clarendonian; Hemph. = Hemphillian; Irving. = Irvingtonian; Ranchol. = Rancholabrean; E = early; M = middle; L = late Palaeolama weddelli, Lama guanicoe, Lama castelnaudi, Vicugna, Vicugna provicugna, and Eulamaops parallelus. Baskin and Thomas (2016) validated two Camelops species: C. hesternus and C. minidokae. Morphological phylogenies have predominantly assigned Camelops to the Lamini clade (Harrison 1979;Honey et al. 1998;Scherer 2013), but recently novel proteomic (Buckley et al. 2019) and genetic (Heintzman et al. 2015) studies indicated a closer connection between this genus and the Camelini. By contrast, no progress has been made to determine the intra-and inter-generic affinities of Aepycamelus. Morphological phylogenies have also neglected the intrageneric relationships of Alforjas, Pleiolama, Procamelus, and Camelops. Our study provides novel insights into these areas of the Camelid tree, based on revision of original materials in museums, a critical assessment of the literature, and an analysis of these data.

Materials
Specimen PIMUZ A/V 4165 is from the Roth Collection at the Palaeontological Museum of the University of Zurich. It was collected in Barranca del Parana, San Nicolas (Buenos Aires province, Argentina) (Roth 1889;Schulthess 1920). The stratigraphic age is Ensenadan (early to middle Pleistocene) (Roth 1889;Schulthess 1920;Cione et al. 2015), which dates approximately from ~ 1.95-1.77 to 0.4 Mya (Verzi et al. 2004;Soibelzon et al. 2008). To investigate the taxonomic and systematic allocation of this fossil, we conducted a phylogenetic analysis of several South and North American taxa. We used Poebrotherium wilsoni, a basal camelid , as the outgroup. For the ingroup, we included members of the Camelinae (see Additional files 1, 2). Several North American taxa (Pleiolama, Camelops, Aepycamelus, Procamelus, and Alforjas), previously scored as genera by Scherer (2013), Honey et al. (1998) and Harrison (1979, are now scored as species. For Pleiolama and Camelops, we included all species currently accepted. We also included four species of Aepycamelus and a single one of the genus Alforjas. All Protolabinae genera (Michenia, Tanymykter, Protolabis) are also in the ingroup.
Most of the terminal taxa are scored from a single, well-preserved specimen. A few taxa, such as Camelops minidokae, Megatylopus matthewi, and Palaeolama mirifica, were scored with more fragmentary material. For Lama and Vicugna, we included only the wild forms (L. guanicoe, V. vicugna). For Camelus, we scored material from the domestic C. bactrianus.
We reviewed previous systematic analyses by Scherer (2013), Honey et al. (1998), Harrison (1979, and Webb (1965). Characters were updated and adapted, as noted. Five characters, not previously used in cladistic analyses, were based on information provided by Webb (1965) and by Honey (2007). In total, we present 17 new characters. We measured the hypsodonty index as per Shockey (1997). We found it to vary greatly, particularly in our sample of V. vicugna, likely due to differences in the level of wear. Therefore, we omitted this character from our matrix. We followed the scoring of Scherer (2009Scherer ( , 2013 for P 3 (char. 12) and P 3 (char. 13) in L. guanicoe and V. vicugna, as she had access to more material from these species and because their states were ambiguous in our sample.

Systematic analysis
The character matrix is available in the Electronic Supplementary Material (Additional file 5). It comprises 49 morphological characters. Eight of these characters (chars. 8, 9, 11, 26, 29, 33, 45, and 47) were ordered. For the ingroup, we examined 18 species and 3 specimens identified at the genus level distributed among 13 Camelinae genera and 3 Protolabinae genera (Additional file 2). We performed a cladistic analysis using parsimony with a traditional search in TNT (Goloboff and Catalano 2016). We ran 6000 replicates with 10 trees saved per replication. Bremer supports were obtained with suboptimal trees by 7 steps. Bootstraps values were obtainedwith a cutoff of 50.

Locality and horizon
Specimen PIMUZ A/V 4165 was collected near the Parana river in the locality of San Nicolas (Barranca del Parana, San Nicolas, province of Buenos Aires, Argentina) (see Roth 1888, Fig. 1). There is no further information on the specimen or its precise stratigraphic position within las Barrancas. Geological age is Ensenadan (Early-to-Middle Pleistocene) according to museum archives and current curatorial staff at Museo de La Plata (Roth 1889;Schulthess 1920;Cione et al. 2015, A. Carlini, personal communication, July 18, 2020. However, there has been no modern revision of the stratigraphy at Roth's collection sites, so these dates must be interpreted with caution. Description PIMUZ A/V 4165 is a nearly complete skull and mandible with all upper and lower teeth except a fragment of right I 3 and left I 3 . Postcranial elements include all metapodials, carpals, tarsals, radio-ulnae, humeri, and tibiae as well as the left femur, a damaged left scapula, and most phalanges. The axial skeleton (pelvis, vertebrae, and ribs) and the patellae are absent.

Skull and mandible
The skull dimensions are similar to those of Lama guanicoe. Length from the posterior edge of the occipital condyle to the anterior end of the rostrum is 31.5 cm (tip broken) and the width is 13.7 cm measured at the greatest breadth of the skull. The skull displays a high degree of flexion at the boundary between the basicranium and palate. The angle of flexion is 14°-it is among the highest  (ZM 17,967), left side, scored rounded (2). b P. coartatus (AMNH 73,306), right side, scored ventrally produced with lateral flare (1). c, d: Char. 43: "Coronoid process at the level of the mandibular condyle in lateral view". c T. brachyodontus (AMNH 36,594), right side, scored angled posteriorly (0). d L. guanicoe (ZM 17,967), left side, scored straight (1). e, f: Char. 40: "Distance between mesial end of the third lower incisors compared to the distance between the mesial end of fourth lower premolars". e C. minidokae (UCMP 38,448), scored distance between I 3 greater than distance between P 4 (1). f P. grandis (UCMP 32,864), scored distance between P 4 greater or equal to distance between I 3 (0). g, h: Char. 39: "Anterior part of the mandible in lateral view". g L. guanicoe (ZM 17,967), left side, scored straight (0). h T. longirostris (CM 2498), left side, scored angled dorsally (1). i Char. 41: "Portion of the mandibular ramus below the mandibular diastema in transverse section". Left: Little narrowing dorsally, ovoid (1). Right: Strongly narrowing dorsally, tear-drop shaped (0) in Lama, according to data provided by Webb (1965). As in all camelids, the rostrum becomes extremely tapered, starting anterior to the orbits.
The skull displays many morphological characters consistent with South American camelids. The zygomatic arch is sigmoid in lateral view, and the maxillary fossa is absent. The nasal bones are strongly retracted, and the anterior edge of the choanae is rounded and situated anteriorly at the level of M 2 . In more basal camelids outside of the Lamini clade, such as Poebrotherium, the nasal bones are longer and extend closer to the anterior edge of the premaxillae, resulting in a smaller opening. In these earlier camelids, the anterior edge of the choanae tends to be more posteriorly positioned and "V-shaped" (Fig. 10).

Teeth
The dental formula is I1/3 C1/1 P2/2 M3/3, with premolars in the third and fourth positions. Among Lamini, this dental formula is also found in L. guanicoe and H. paradoxa. PIMUZ A/V 4165 retains the P 3 , which is absent in other camelids, such as Camelops. "Llama buttresses" (protostylids and parastylids) are well developed. Upper canines and incisors are small, compared to the great size which they reach in some males of extant South American camelids.
Anterior stylids (protostylids and parastylids) on M 3 are prominent, and the anterior fossette on M 1 is still present, making this specimen correspond to the "wear stage 3" of Breyer's classification (Breyer 1977). The molar lophs and lophids are rounded in occlusal view. "Ribs" (labial styles) on the upper molars are well developed and P 4 is triangular, with a fossetid on the distal lobe.

Limbs
Compared to data on South American camelids from Scherer (2009, p. 163-166, figs. 18-19, p. 68-70), limb bones of this specimen are longer than in Vicugna vicugna. Most limb bones are also longer than Lama guanicoe, except its metacarpals which fall in the upper values of L. guanicoe (measurements of PIMUZ A/V 4165 in Additional file 1). Compared to Palaeolama and Hemiauchenia, metapodial lengths overlap with Palaeolama and are smaller than in Hemiauchenia. Tibiae and radio-ulnae lengths both overlap with H. paradoxa and P. mirifica, and, respectively, with P. major and P. weddelli. Stylopodials are shorter than in Palaeolama and Hemiauchenia.
Metacarpals are more gracile than in Palaeolama. Metatarsal gracility falls within the lower values of Palaeolama major. Metacarpals and metatarsals are more robust than in Vicugna and more in the range of Lama, H. paradoxa, and H. macrocephala.
In the hind limbs, metatarsals are smaller than the tibiae and femora. Femora are shorter than the tibiae, as in V. vicugna, H. paradoxa, and H. macrocephala (char.  (2013)). In the forelimbs, humeri are the shortest bones, and radio-ulnae are the longest. This is in contrast to some Lamini, including extant forms and Palaeolama, whose humeri are longer or of comparable length than their metacarpals (char. 28 in Scherer (2013)) (Figs. 11, 12).

Results
The cladistic analysis resulted in one most parsimonious tree (Fig. 12) with length 149 (CI = 0.423, RI = 0.675, RC = 0.286). Synapomorphies are listed in Additional file 6. Both Bootstrap values and Bremer supports are low overall.
PIMUZ A/V 4165 is positioned within South American Lamini and is the sister taxon to the clade formed by Lama guanicoe and Vicugna vicugna (node 52). The synapomorphy that holds these three taxa together at node 47 is a small postglenoid process (char. 35). PIMUZ A/V 4165 has a single autapomorphy: a lower canine that is close to I 3 (< 1 cm) (char. 7).
Well-developed protostylids and parastylids ("llama buttresses") (char. 15) are the single synapomorphy of the clade formed by H. paradoxa, PIMUZ A/V 4165, L. guanicoe, and V. vicugna (node 45). Hemiauchenia appears as polyphyletic. Palaeolama (node 48) is monophyletic; its synapomorphies are a quadrangular P 4 with fossetids in the mesial and distal lobes (char. 14), and triangular labial lophids (char. 16). The clade formed by L. guanicoe and V. vicugna (node 52) is held together by the absence of P 3 and P 3 (chars. 12 and 13), and the foramen ovale and median lacerate foramen are completely separated by a thin spine of the alisphenoid (char. 33). The confluence of these foramina is an autapomorphy of L. guanicoe. Pleiolama is not monophyletic, appearing at two different positions within Lamini: P. vera is the basal-most Lamini taxon (node 37), while P. mckennai is the most basal taxon in the clade formed mostly by North American Lamini from the Miocene (node 35). The genera Aepycamelus and Procamelus are, respectively, polyphyletic and paraphyletic. A. elrodi and A. alexandrae are held together by four synapomorphies (node 49); A. bradyi forms a clade with Procamelus sp. (node 40) and A. robustus is more basal (node 34). Procamelus also appears in two different places, but close to each other: nodes 32 and 40.
Camelops (node 51) is monophyletic and placed within the Camelini (node 43). This genus is united by the absence of P 1 (char. 8), anterior edge of the choanae U-shaped (char. 30), ethmoid foramen positioned in the fossa formed by a division of the lateral pterygoid muscle (char. 32), and the ovale and median lacerate foramina are completely separated by a thin spine of the alisphenoid (char. 33).
Camelinae (node 39) are monophyletic with three synapomorphies supporting the group: caniniform P 1 and P 1 (char. 8 and 9), and absence of P 2 (char. 10). Alforjas taylori is a basal Camelinae, sister taxon to the clade formed by the 'tribes' Camelini and Lamini. Our definitions of Camelini and Lamini differ from those of Honey et al. (1998). We interpret the first major division within the Camelinae (node 38) as the split between these two clades.
The three synapomorphies uniting the Camelini (node 43) are the absence of P 3 (char. 13), well-developed upper molar labial styles ("ribs") (char. 17), and a lacrimal vacuity bordered by four bones, with the lacrimal border greatly reduced and the frontal border greatly enlarged (char. 25). The lacrimal vacuity bordered by two bones, frontal and maxillary, is an autapomorphy of Camelus bactrianus. Megatylopus and Camelus bactrianus (node 42) are united by a single synapomorphy: the absence of the premaxillary notch (char. 22).
The Protolabinae (composed of Tanymykter, Michenia, and Protolabis) are paraphyletic in our results. The genus Tanymykter (node 28) is monophyletic and supported by two synapomorphies.

Discussion
Most of the terminal taxa were scored on the skull of a single specimen; therefore, our results could be altered by a larger sampling for each species and the addition of postcranial characters.
PIMUZ A/V 4165 appears most closely related to L. guanicoe and V. vicugna (node 47), to the exclusion of H. paradoxa. We found the postglenoid process (char. 35) of H. paradoxa to be large, whereas a small postglenoid process unites PIMUZ A/V 4165 with L. guanicoe and V. vicugna. A lower canine closer to I 3 (< 1 cm) is the autapomorphy for PIMUZ A/V 4165. In H. paradoxa, L. guanicoe, and V. vicugna, the lower canine is positioned far from I 3 (≥ 1 cm) (char. 7).
We found important differences between the postcranials of PIMUZ A/V 4165 and those of L. guanicoe and V. vicugna. Among these, PIMUZ A/V 4165 has longer limb bones than V. vicugna and longer or among the largest sizes of L. guanicoe (Scherer 2009). In V. vicugna, metacarpals are shorter than metatarsals (Scherer 2013), whereas in PIMUZ A/V 4165, metacarpals are slightly longer than metatarsals. Metacarpals are longer than the humeri in PIMUZ A/V 4165, in contrast to L. guanicoe whose metacarpals are shorter than the humeri (Scherer 2009, p. 201), and V. vicugna whose metacarpals and humeri lengths are comparable (Scherer 2009, p. 232).
The postcranial characters observable on PIMUZ A/V 4165 are most similar to H. paradoxa in Scherer's (2013) matrix (see Additional file 7). For example, as in PIMUZ A/V 4165, metacarpals are longer than humeri in H. paradoxa. However, an analysis using Scherer's (2013) matrix results in a polytomy with PIMUZ A/V 4165 within the Lamini (see Additional file 7). Furthermore, compared to data from Scherer (2009), PIMUZ A/V 4165 has shorter metatarsals and humeri than H. paradoxa. Although the parameters of our study cannot confirm it, the fact that the morphology of PIMUZ A/V 4165 is not consistent with any single species could indicate that it belongs to a new species.
The monophyly of the Protolabinae recovered by Honey and Taylor (1978), , Honey et al. (1998) and Scherer (2013) is not supported here. The species previously assigned to this group display much morphological disparity. For example, Michenia agatensis and Tanymykter retain I 1 and I 2 (chars. 1 and 2), while Protolabis coartatus has lost these teeth. We found these absences (chars. 1 and 2) to be synapomorphies uniting P. coartatus to the Camelinae (node 30), whereas in Scherer (2013), these were synapomorphies in her definition of the Camelinae. Here, the caniniform P 1 and P 1 (chars. 8 and 9) and the absence of P 2 (char. 10) are also synapomorphies of the Camelinae (node 39). Of these, the synapomorphies of characters 8 and 9 are in agreement with Honey et al. (1998), and the synapomorphy of character 10 is in agreement with Harrison (1979).
Our phylogeny places Alforjas taylori in the most basal position among the Camelinae-this is in significant contrast to previous phylogenies. Harrison (1979 and Scherer (2013) placed this species in a more derived position within Lamini. In Scherer (2013), this was partly due to postcranial characters. Here and in Scherer (2013), the presence of the protostylids and parastylids ("llama buttresses") (char. 15) on the lower molars is not a synapomorphy for the Lamini (node 37). In contrast, Harrison (1979 found this feature as a synapomorphy for the Lamini. In Harrison (1979 and Honey et al. (1998), Alforjas and Camelops were united partly by a similar degree of hypsodonty. In these previous works and in Scherer (2013), Alforjas and Camelops were placed within Lamini and shared with them dorsally arched nasal bones in the transverse section (char. 23). We agree with this being a synapomorphy for Lamini (node 37), but we were unable to score it on A. taylori due to the poor preservation of the nasal bones. In C. hesternus and C. cf. hesternus, we found this feature to be dorso-ventrally flattened.
In our results, Camelops (node 51) is included in the Camelini (node 43). This is in agreement with recent proteomic and genetic studies which defined Camelops as more closely related to Camelini than to the Lamini (Buckley et al., 2019;Heintzman et al., 2015). Harrison (1979 and Scherer (2013) found rounded canines (char. 5) to be a synapomorphy of the Camelini, whereas we found it to be a synapomorphy of the clade composed of Aepycamelus and Procamelus (node 34). Procamelus appeared as a basal Camelinae in Scherer (2013) and Harrison (1979). Honey et al. (1998) placed it within the Camelini. Here, as in , Procamelus is more derived and placed within the Lamini. Aepycamelus is also positioned within the Lamini, as in Scherer (2013) and Honey et al. (1998). However, these genera are not monophyletic in our results, which indicates that further taxonomic revisions are necessary both taxa. We agree with observations by Honey et al. (1998) on the ambiguous status of Aepycamelus. We found important morphological differences between the different species of Aepycamelus. For instance, P 2 (char. 10) is absent in A. robustus, whereas it is present in A. alexandrae, A. bradyi, and A. elrodi.
Scherer (2013) did not find Hemiauchenia to be monophyletic and we agree on this point. However, we also find a polyphyletic separation between H. paradoxa and H. macrocephala. The monophyly of Palaeolama (node 48) was also recovered in Scherer (2013), and in both studies, a quadrangular P 4 , with fossetids in the mesial and distal lobes (char. 14), is a synapomorphy for this genus. Here, Palaeolama is closely related to the genera Hemiauchenia, Vicugna, and Lama, whereas in Scherer (2013), Palaeolama was in a more basal position.

Dental characters
When present, P 1 , P 1, and P 2 (chars. 8, 9, and 11) are either premolariform or caniniform. We ordered these characters, assuming a linear evolution from a plesiomorphic premolariform state towards a caniniform state, with eventual loss in more derived taxa. At multiple points in our results, the states for these characters did not follow this linear path. For example, one of the synapomorphies at node 34 is a change from absent to a caniniform P 2 (char. 11). Similarly, on node 33, P 2 changes from caniniform to premolariform.