Early archosauromorph remains from the Permo-Triassic Buena Vista Formation of north-eastern Uruguay

The Permo-Triassic archosauromorph record is crucial to understand the impact of the Permo-Triassic mass extinction on the early evolution of the group and its subsequent dominance in Mesozoic terrestrial ecosystems. However, the Permo-Triassic archosauromorph record is still very poor in most continents and hampers the identification of global macroevolutionary patterns. Here we describe cranial and postcranial bones from the Permo-Triassic Buena Vista Formation of northeastern Uruguay that contribute to increase the meagre early archosauromorph record from South America. A basioccipital fused to both partial exoccipitals and three cervical vertebrae are assigned to Archosauromorpha based on apomorphies or a unique combination of characters. The archosauromorph remains of the Buena Vista Formation probably represent a multi-taxonomic assemblage composed of non-archosauriform archosauromorphs and a ‘proterosuchid-grade’ animal. This assemblage does not contribute in the discussion of a Late Permian or Early Triassic age for the Buena Vista Formation, but reinforces the broad palaeobiogeographic distribution of ‘proterosuchid grade’ diapsids in Permo-Triassic beds worldwide.


Geological and palaeontological setting
The Buena Vista Formation crops out in north-eastern Uruguay and consists of red-brownish sandstones, intercalated with thin layers of red-brownish mudstones and intraformational conglomerates deposited under continental fluvial conditions (Bossi & Navarro, 1991;Goso et al., 2001;. Most of the fossils collected from the Buena Vista Formation are found in the intraformational conglomerates close to the Colonia Orozco town and correspond to isolated bones to ocassionally articulated partial skeletons (Fig. 1). The tetrapod fossil content of the Buena Vista Formation represents the Colonia Orozco Local Fauna and includes laidleriid (Uruyiella liminea Piñeiro, Marsicano & Lorenzo, 2007), mastodonsaurid, rhinesuchid-like and dvinosaurian temnospondyls (Marsicano, Perea & Ubilla, 2000;Piñeiro, 2004;Piñeiro, Marsicano & Damiani, 2007;Piñeiro, Ramos & Marsicano, 2012), procolophonoid parareptiles (i.e., Pintosaurus magnidentis Piñeiro, Rojas & Ubilla, 2004;Piñeiro, 2004), probable varanopid and sphenacodontid synapsids Piñeiro et al., 2013), and basal archosauromorphs Piñeiro, 2004) (present study). The age of the Buena Vista Formation is poorly constrained because of the absence of index taxa and the presence of taxa that are documented in either Late Permian or earliest Triassic assemblages. Therefore, the age of this formation has been substantially debated. The Buena Vista Formation has been considered a lateral equivalent of the Sanga do Cabral Formation of southern Brazil based on lithostratigraphic similarities (Andreis, Bossi & Montardo, 1980;Bossi & Navarro, 1991). The Brazilian unit is considered late Induan-early Olenekian in age because of the presence of the index taxon Procolophon (Dias-da-Silva, Modesto & Schultz, 2006) and, as a result, the same age has been assigned to the Buena Vista Formation (Bossi & Navarro, 1991). However, subsequent authors have suggested an older age for the Buena Vista Formation, being closer to the Permo-Triassic boundary or even within the Late Permian based on its tetrapod fossil content Piñeiro, 2004;Piñeiro, Rojas & Ubilla, 2004;Piñeiro, Marsicano & Lorenzo, 2007;Piñeiro, Marsicano & Damiani, 2007;Piñeiro, Ramos & Marsicano, 2012). In particular, the description of varanopid synapsids would favour a Permian age , but recent authors have casted doubts on these assignments and concluded that there is no compelling evidence to support a Permian age over an Early Triassic one (Dias-da-Silva, Modesto & Schultz, 2006). Here, we will consider the Buena Vista Formation as Permo-Triassic in age, taking into account recent studies that placed the Colonia Orozco Local Fauna as a transitional assemblage that could contain the Permo-Triassic boundary (see Piñeiro, Ramos & Marsicano, 2012
The lateral surface of the basioccipital is subdivided into dorsolaterally and lateroventrally facing surfaces. Both surfaces meet each other in an obtuse angle in posterior view at the apex of the lateral flange of the bone. The dorsolaterally facing surface is flat and probably participated of the medial wall of the metotic foramen (Figs. 2C and 2E: wmf). The ventrolaterally facing surface is damaged on the left side of the bone ( Fig. 3E: ds), but well preserved on the right side. The ventrolaterally facing surface has a complex topology and is subdivided by a diagonal, posteroventrally-to-anterodorsally oriented ridge ( Fig. 2: dr). The facet for reception of the ventral ramus of the opisthotic is situated posterodorsally to this ridge ( Fig. 2C: fvro). This facet is posterodorsally-to-anteroventrally slightly concave. The surface anteroventral to the diagonal ridge is more deeply anteroposteriorly concave than the facet for the ventral ramus of the opisthotic and is delimited anteriorly by the facet for the parabasisphenoid (Fig. 2: fpbs). The presence of smooth cortical bone on this deeply concave surface indicates that probably it was a non-articulating surface and potentially might have been part of the medial wall of the passage of the pseudolagenar recess ( Fig. 2: mwpr). The pseudolagenar recess is present in Prolacerta broomi, Euparkeria capensis, several proterosuchian-grade archosauriforms and the poposauroid Xilousuchus sapingensis Wu, 1981(Gower & Sennikov, 1996. The anterior surface of the basioccipital has a slightly transversely convex facet for articulation with the parabasisphenoid (Fig. 2: fpbs). This articular facet extends also onto the anterodorsal surface of the basioccipital, immediately lateral to the floor of the endocranial cavity. The floor of the endocranial cavity is flat and has an anteroposteriorly long median longitudinal ridge, which is restricted to the anterior half of the basioccipital (Figs. 2B and 2E: mlr), resembling the condition in some procolophonids (e.g., Leptopleuron lacertinum (Spencer, 2000)), synapsids (e.g., Haptodus garnettensis: (Laurin, 1993)) and diapsids (e.g., Youngina capensis: (Gardner et al., 2010); Gephyrosaurus bridensis: (Evans, 1980)). The floor of the endocranial cavity of Prolacerta broomi has a pair of longitudinal ridges that delimit a shallow, median groove along most of the dorsal surface of the basioccipital (BP/1/2675).
The absence of suture between the right exoccipital and basioccipital precludes determining if the exoccipitals contact each other on the floor of the endocranial cavity. The foramen/foramina for the exit of the hypoglossal and glossopharyngeal cranial nerves (CN XI-XII) are not preserved.

Table 2 Measurements of the anterior (FC-DPV 2640), middle-posterior (FC-DPV 2639), and middleposterior (FC-DPV 2637) cervical vertebrae in millimeters.
Values between brackets indicate incomplete measurements and between squared brackets indicate estimated measurements. The length along the zygapophyses is the maximum anteroposterior length between the anterior tips of the prezygapophyses and the posterior tips of the postzygapophyses. Maximum deviation of the digital caliper is 0.02 mm but measurements were rounded to the nearest 0.1 millimeter.

FC-DPV2640
FC Anterior cervical vertebra. FC-DPV 2640 (Figs. 5A-5C; Table 2) is interpreted as an anterior cervical vertebra because of its strong anteroposterior elongation and a facet for articulation with the rib (only the base of the left structure is preserved) placed next to the anterior margin of the neural arch. The anterior end of the centrum is damaged, but it seems to have been strongly bevelled and anteroventrally facing. If this condition is not an artefact due to damage, FC-DPV 2640 may correspond to an axis because it has enough room to receive the intercentrum of the axis. The posterior surface of the centrum is concave and seems to be slightly bevelled, possibly to receive a small intercentrum. The vertebra is possibly not notochordal. The centrum is approximately 3.6 times longer than tall, a ratio that closely resembles the condition in the axis of tanystropheids (e.g., Tanystropheus longobardicus Bassani, 1886: (Nosotti, 2007; Amotosaurus rotfeldensis Fraser & Rieppel, 2006) and the third and fourth cervical vertebrae of moderately long-necked basal archosauromorphs, such as Protorosaurus speneri (BSPG 1995 I 5, cast of WMSN P47361), Prolacerta broomi (BP/1/2675) and Macrocnemus bassanii (Nopcsa, 1930) Lane, 1945(Reisz, 1981, Gephyrosaurus bridensis (Evans, 1981), Trilophosaurus buettneri (Spielmann et al., 2008), Prolacerta broomi (BP/1/2675) and Proterosuchus alexanderi (NMQR 1484). Only the base of the left facet for articulation with the rib is preserved and is restricted to the anteroventral portion of the centrum. The neurocentral suture is completely closed. The zygapophyses lack their distal ends, but their preserved portions indicate that they were anteroposteriorly long, laterally divergent and sub-horizontal. As a result, the distal tips of the zygapophyses are well separated from the median line, resembling the condition in the anterior and middle cervicals of other basal archosauromorphs (e.g Trilophosaurus buettneri: (Gregory, 1945;Spielmann et al., 2008)). The neural arch has a very shallow depression lateral to the base of the neural spine (Figs. 5A and 5B: d), as occurs in Prolacerta broomi (BP/1/2675) and several other basal archosauromorphs. By contrast, the middle and posterior cervical vertebrae of at least some specimens of Proterosuchus alexanderi have a better defined and deeper, sub-circular fossa lateral to the base of the neural spine (NMQR 1484). The neural spine is mostly complete, but its dorsal margin is damaged where it becomes very thin transversely (Figs. 5A and 5B: ns). As a result, it is interpreted that the neural spine should not have been much taller and preserves it general shape. The neural spine is dorsoventrally short and strongly elongated anteroposteriorly, as occurs in Protorosaurus speneri (BSPG 1995 I 5, cast of WMSN P47361), Prolacerta broomi (BP/1/2675), Amotosaurus rotfeldensis (SMNS 50830) and Macrocnemus bassanii (PIMUZ T2472, T4355, T4822), but contrasting with the taller and anteroposteriorly shorter neural spine of Proterosuchus alexanderi (NMQR 1484).
In the neural arch, a posterior centrodiapophyseal lamina delimits a central infradiapophyseal fossa below the base of the transverse process ( Fig. 6: pcdl). In addition, a tuberosity runs from the base of the transverse process towards the base of the postzygapophysis, but it does not reach the latter structure. This tuberosity and the posterior centrodiapophyseal lamina delimit a shallow subtriangular depression that is topologically equivalent to a postzygapophyseal centrodiapophyseal fossa. There are no anterior centrodiapophyseal and prezygodiapophyseal laminae in the neural arch, which may be a result of the relatively anterior position of the vertebra in the cervical series. The zygapophyses are horizontal and anteroposteriorly short, but extend slightly beyond the margins of the anterior and posterior articular facets of the centrum, respectively. The prezygapophyses are anterolaterally directed and, as a result, their distal tips are well separated from the median line ( Fig. 6: prz), as occurs in the cervico-dorsal vertebrae of Macrocnemus bessanii (PIMUZ T482), Tanystropheus longobardicus (Wild, 1973) and Trilophosaurus buettneri (Spielmann et al., 2008). The articular surfaces of the zygapophyses are damaged and it is not possible to determine their morphology. A shallow and poorly defined, circular depression is present laterally to the base of the neural spine ( Fig. 6: d), as occurs in at least some specimens of Proterosuchus alexanderi (NMQR 1484). The neural spine is moderately low and strongly anteroposteriorly elongated, being considerably anteroposteriorly longer than tall (Fig. 6: ns), closely resembling the condition in Protorosaurus speneri (BSPG 1995 I 5), Macrocnemus bessanii (PIMUZ T4822) and Prolacerta broomi (BP/1/2675). By contrast, in Mesosuchus browni (SAM-PK-5882), Trilophosaurus buettneri (Spielmann et al., 2008), Proterosuchus fergusi Broom, 1903(BSPG 1934GHG 231), Sarmatosuchus otschevi (PIN 2865/13-19), Erythrosuchus africanus (NHMUK R3592), Garjainia prima (PIN 2394/5-13, 5-16) and Euparkeria capensis (SAM-PK-586) the neural spines are taller than long. The neural spine has an anterior overhang that extends anteriorly beyond the base of the spine (Fig. 6: ao), as occurs in Protorosaurus speneri (BSPG 1995 I 5), Macrocnemus bessanii (PIMUZ T4822), Trilophosaurus buettneri (Spielmann et al., 2008) and Prolacerta broomi (BP/1/2675). The distal margin of the neural spine has a low transverse thickening (Fig. 6: dt), but it does not form a spine table or a mammillary process. The same thickening on the distal margin of the neural spine is present in several other long-necked archosauromorphs (e.g., Macrocnemus bessanii: PIMUZ T4822; Prolacerta broomi: BP/1/2675). The pre-and postspinal fossae are deep and transversely wide ( Fig. 6: posf, prsf). The prespinal fossa is restricted to the base of the neural spine and the postspinal fossa extends onto most of the posterior surface of the spine, as usually occurs in other basal archsoauromorphs (e.g., Prolacerta broomi: BP/1/2675).
Middle-posterior cervical vertebra. FC-DPV 2637 (Figs. 5D-5H; Table 2) belongs to a middle or posterior cervical vertebra because the parapophyses are situated slightly above the mid-height of the centrum, adjacent to its anterior margin (Fig. 5E: pa), and the neural spine is anteroposteriorly short (Figs. 5E and 5F: ns). This vertebra is well-preserved, but moderately squeezed posteroventrally to the right side (Fig. 5H), the posterior articular facet of the centrum is damaged and most of the prezygapophyses, right diapophysis and neural spine are missing. The centrum is amphicoelous and apparently not notochordal. The centrum length represents 1.7 times the height of its anterior articular facet, being proportionally shorter than FC-DPV 2639 and resembling the ratio present in the middle-posterior cervical vertebrae of several basal archosauromorphs (e.g., Aenigmastropheus parringtoni: UMZC T836; Eorasaurus olsoni: PIN 156/109; Trilophosaurus buettneri: (Spielmann et al., 2008); Proterosuchus alexanderi: NMQR 1484). The ventral surface of the centrum is transversely convex and lacks a median ventral keel. The centrum is slightly transversely compressed at mid-length. The anterior articular facet of the centrum is transversely broader than tall. The contour of the posterior facet cannot be determined because of damaging ( Fig. 5H: pfc). The parapophyses are situated on laterally projected peduncles (Figs. 5D, 5E, 5G and 5H: pa). The peduncle of the parapophysis has a moderately deep depression on its ventral surface. The facet of the parapophysis is semi-circular, with a mostly straight anterior margin, and mainly laterally facing, with a low anteroventral component. A sub-horizontal ridge extends posteriorly from the base of the parapophysis to the lateral surface of the centrum, but it does not reach the level of mid-length of the centrum (Figs. 5D and 5E: arrow). A similar ridge is also present in FC-DPV 2640 (Figs. 5A and 5B) and the middle and posterior cervical vertebrae of other basal archosauromorphs, such as Macrocnemus bassanii (PIMUZ T4822), Tanystropheus longobardicus (PIMUZ T2818), Eorasaurus olsoni (PIN 156/108,109) and Garjainia prima (PIN 2394/5-11, 5-13). The lateral surface of the centrum lacks a lateral fossa and the neurocentral suture is completely closed.
The anterior cervical vertebra FC-DPV 2640 and middle-posterior cervical vertebra FC-DPV 2639 are assigned to an archosauromorph diapsid because of the following combination of characters: probable non-notochordal and anteroposteriorly elongated centra, a sub-horizontal ridge on the lateral surface of the centra, a shallow fossa immediately lateral to the base of the neural spines, and neural spines considerably anteroposteriorly longer than tall (Fig. 5). In addition, the centrum of FC-DPV 2639 is parallelogram-shaped in lateral view, a character that was found as a synapomorphy of Archosauromorpha (Ezcurra, Scheyer & Butler, 2014), and has an anterior overhang and a transversely thickened distal margin on the neural spine, features that occur together in the basal archosauromorphs Macrocnemus bessanii (PIMUZ T4822) and Prolacerta broomi (BP/1/2675). These anterior and middle-posterior cervical vertebrae differ from those of "pelycosaur" synapsids and araeoscelidian diapsids in the presence of a probable non-notochordal centrum and a lower and longer neural spine (Ezcurra, Scheyer & Butler, 2014). Among long-necked basal archosauromorphs, FC-DPV 2637 and FC-DPV 2640 differ from Prolacerta broomi and Trilophosaurus buettneri in the presence of a low longitudinal lateral crest that runs posteriorly from the base of the facet for articulation with the rib, from Macrocnemus bessanii in the absence of epipophyses (PIMUZ T4822), and from other tanystropheids in the presence of a proportionally anteroposteriorly shorter centrum (e.g. Amotosaurus rotfeldensis: SMNS 50830). Although the neural spines have damaged distal margins, they seem to have been dorsoventrally short. As a result, they may have differed from Protorosaurus speneri (BSPG 1995 I 5), which has tall neural spines. The combination of characters observed in FC-DPV 2640 is consistent with that present in basal archosauromorphs, such as Prolacerta broomi (BP/1/2675), but FC-DPV 2639 differs from this species in the presence of proportionally anteroposteriorly shorter centrum, and a less developed anterior overhang on the neural spine. FC-DPV 2639 and FC-DPV 2640 differ from the protorosaur cervical vertebra described by Dias-da-Silva (1998) from the Sanga do Cabral Formation in being considerably anteroposteriorly shorter. However, these differences could be due to the position of the vertebrae in the cervical series and the possibility that they belong to closely related species cannot be ruled out. Finally, the probable presence of a single facet for articulation with the cervical rib is a feature shared with non-archosauromorph diapsids (e.g., Gephyrosaurus bridensis: (Evans, 1981); Planocephalosaurus robinsonae: (Fraser & Walkden, 1984)) and tanystropheids (e.g., Amotosaurus rotfeldensis: SMNS 50830; Tanystropheus longobardicus: (Wild, 1973;Nosotti, 2007)). By contrast, more derived archosauromorphs have distinct parapophyses and diapophyses in the postaxial cervicals (e.g., Mesosuchus browni: (Dilkes, 1998); Trilophosaurus buettneri: (Spielmann et al., 2008); Prolacerta: BP/1/2675). Accordingly, FC-DPV 2639 and FC-DPV 2640 are interpreted as indeterminate basal archosauromorphs, but they might belong to a very basal member of the clade because of the presence of a single facet for the rib.
The middle-posterior cervical vertebra FC-DPV 2637 has a series of characters that were found as synapomorphies of Archosauromorpha or less inclusive clades within the group by Ezcurra, Scheyer & Butler (2014): a trapezoidal and probable non-notochordal centrum, and anterior centrodiapophyseal, posterior centrodiapophyseal, prezygodiapophyseal and postzygodiapophyseal laminae on the neural arch (Figs. 5D and 5E). The combination of characters observed in FC-DPV 2637 resembles that present in several disparate basal archosauromorphs (e.g., Protorosaurus speneri: BSPG 1995 I 5; Tanystropheus longobardicus: PIMUZ T2817; Spinosuchus caseanus: Spielmann et al., 2009). Nevertheless, FC-DPV 2637 differs from Prolacerta broomi and Proterosuchus fergusi in the presence of a posterior centrodiapophyseal lamina. FC-DPV 2637 cannot be properly compared with the protorosaur cervical vertebra from the Sanga do Cabral Formation (Dias-da-Silva, 1998) because the latter belongs to a more anterior element in the axial series. As a result, FC-DPV 2637 is interpreted as an indeterminate basal archosauromorph, but distinct from Prolacerta broomi and proterosuchids.

Discussion
The presence of archosauromorphs in the Permo-Triassic Buena Vista Formation was previously reported by Piñeiro (2002) and , but no detailed description or taxonomic discussion have been provided so far. The cranial and postcranial remains described here increase the meagre archosauromorph record in Permo-Triassic rocks of South America.
The partial braincase with resemblances to Proterosuchus is particularly interesting because it probably reinforces the broad palaeobiogeographic distribution of proterosuchids during Permo-Triassic times (i.e., European Russia, China, South Africa and possibly Australia) (Ezcurra, Butler & Gower, 2013). However, we need to be cautious about this hypothesis because the specimen shows some differences that could not be properly compared with the Permian representatives of the group, such as the possible broad contribution of the basioccipital to the ventral margin of the foramen magnum (Fig. 4). The complete skull length of FC-DPV 2641 is estimated between 200-250 mm based on linear regressions of the total length of the skull versus the width and height of the occipital condyle, respectively, of a series of proterosuchid skulls from the Lystrosaurus Assemblage Zone of South Africa (N = 4: NMQR 880, 1484, BSPG 1934; occipital condyle width: y = 0.0514x-0.02115, R 2 = 0.96; occipital condyle height: y = 0.0295x + 2.8892, R 2 = 0.80). The skull length range recovered for FC-DPV 2641 falls in the 4th to 18th percentile of the South African proterosuchid sample (N = 14, total skull length ranges from 177 to 477 mm; Ezcurra & Butler, 2014). The presence of a medium-sized basal archosauromorph in the Permo-Triassic of South America is not unexpected because of the presence of relatively large proterosuchids in the latest Permian of Russia (Archosaurus rossicus) and relatively large basal archosauromorphs in the earliest Triassic of Antarctica (Smith et al., 2011), South Africa (Proterosuchus fergusi) and China ('Chasmatosaurus' yuani).
The archosauromorph partial braincase described here belongs to an individual considerably larger than those of the cervical vertebrae. Among the postcranial bones, FC-DPV 2637 pertained to an animal larger than that represented by FC-DPV 2639, 2640 and the latter two vertebrae are similar in size to each other (Table 2: compare transverse width of the centra). Thus, the archosauromorph bones described here should have belonged to multiple individuals. Furthermore, the simultaneous occurrence of a proterosuchid-like partial braincase and a middle-posterior cervical vertebra with clear differences with proterosuchids support the hypothesis of a multi-taxonomic archosauromorph assemblage. The archosauromorph record of the Buena Vista Formation seems to bolster a Permo-Triassic age for the unit, as previously concluded by Piñeiro, Ramos & Marsicano (2012). Unfortunately, this archosauromorph assemblage does not help in the debate of a Permian or Triassic age for this unit because both basal archosauromorphs and early archosauriforms (e.g., proterosuchids) are present across the Permo-Triassic boundary (Ezcurra, Scheyer & Butler, 2014