A small podocnemidoid (Pleurodira, Pelomedusoides) from the Late Cretaceous of Brazil, and the innervation and carotid circulation of side‐necked turtles

Pleurodires are less diverse than cryptodires, together forming the two major lineages of crown turtles. Their fossil record, however, is rich. A particularly large number of fossil pleurodires, many belonging to the Podocnemidoidae, has been recovered from the Upper Cretaceous Bauru Group outcrops of south‐central Brazil. Herein we describe an additional pleurodire from this region, Amabilis uchoensis gen. et sp. nov., based on a partially preserved skull. A. uchoensis is recognized as belonging to the Podocnemidoidae by the small entrance to the antrum postoticum and completely developed cavum pterygoidei, being unique among other non‐podocnemidid podocnemidoids for its medially open fenestra postotica and absent basioccipital–opisthotic contact. Our cladistic analysis places A. uchoensis as a sister taxon to Hamadachelys + other podocnemidoids. We further explore the neuroanatomy of side‐necked turtles with the aid of micro‐computed tomography of specimens of the main pleurodiran lineages. Our data shed light on the different carotid circulation patterns in pleurodires, and we propose new phylogenetic characters to describe the neuroanatomical variation of the group. Optimization of these characters shows two independent acquisitions of a foramen for the palatal branch of the carotid in chelids and podocnemidoids, and a unique loss of the vidian nerve canal in chelids, in turtles in general.

L I V I N G pleurodires, 93 species of which are currently recognized, account for nearly one-quarter of extant turtle diversity (TTWG 2017), with all remaining extant turtles being cryptodires. Extant pleurodires are restricted to freshwater environments of southern landmasses, namely South America, Africa, Madagascar and Australasia (de la Fuente et al. 2014). Today, three pleurodire clades are recognized: Chelidae, Pelomedusidae and Podocnemidae. However, the fossil record of the group indicates that pleurodire diversity (both in terms of lineages and ecological adaptations) was higher in the past. For instance, by the end of the Early Cretaceous there were at least four established lineages (Romano et al. 2014;Vlachos et al. 2018) and the group included at least two independent invasions of littoral/marine environments (Rabi et al. 2012;Ferreira et al. 2015Ferreira et al. , 2018aJoyce et al. 2016). Additionally, a much broader distribution, reaching all continents except Antarctica and central and northern Asia, can be inferred from the fossil record (Gaffney et al. 2011;Ferreira et al. 2018a).
Although the fossil record of pleurodire cranial remains is relatively rich (e.g. Gaffney et al. 2006Gaffney et al. , 2011Ferreira et al. 2018b), little attention has been given to their carotid and innervation patterns compared with other turtle groups (e.g. Jamniczky et al. 2006;Jamniczky & Russell 2007;Rabi et al. 2013;Joyce et al. 2018;Rollot et al. 2018;Raselli & Anquetin 2019). The carotid system in pleurodires is a matter of high systematic interest, given that a notable variation in position for the related foramina has been recognized, especially among fossils (Gaffney et al. 2006;Miyashita 2013), and also because the patterns of arterial circulation appear to be conservative at the family level in turtles (e.g. McDowell 1961;Albrecht 1967Albrecht , 1976Jamniczky & Russell 2007;Jamniczky 2008). Whereas the position of the foramina related to the entrance of the carotids in the braincase is more commonly taken into account (e.g. Gaffney et al. 2006;Sterli & de la Fuente 2010;Sterli et al. 2010), very little information is known regarding their size, trajectory and development inside the pleurodire skull. These are commonly extrapolated from a small sample of taxa (e.g. Albrecht 1976;Gaffney 1979;Sterli & de la Fuente 2010;Miyashita 2013;Werneburg & Maier 2019), hindering more thorough interpretation, if the topography of canals and foramina is actually not conserved across the pleurodiran lineage. Recent efforts using non-invasive techniques such as micro-computed tomography (lCT) have shed light on these patterns for turtles in general (e.g. Paulina-Carabajal et al. 2013, 2017Evers & Benson 2019;Raselli & Anquetin 2019;Evers et al. 2019a), but to date these have not been applied to a comprehensive sample of pleurodires.
Here, we present a new podocnemidoid turtle from the Late Cretaceous Bauru Group, represented by a single skull from the São Jos e do Rio Preto Formation. We describe its osteology and neuroanatomy, and include the taxon in a phylogenetic analysis with additional characters based on new lCT data. Further, we investigate the morphological variation related to the carotid arterial circulation and cranial nerves in Pleurodira, which traditionally have been considered to follow a consistent pattern across most representatives (e.g. Sterli & de la Fuente 2010;Sterli et al. 2010). We conclude that they are not as conservative as previously thought, having substantive phylogenetic value.

GEOLOGICAL SETTING
The Bauru Basin consists of a large (c. 370 000 km 2 ) depression developed during the Late Cretaceous in the south-eastern portion of the South American Plate. Filled between the Coniacian and Maastrichtian (Fernandes & Coimbra 1996, 2000, the basin has one of the most detailed Cretaceous fossil records of Brazil (see Menegazzo et al. 2016). Within the basin, the São Jos e do Rio Preto Formation crops out in topographic highs of the homonymous region. This unit corresponds essentially to immature sand deposits, often conglomerates, accumulated in bars and fluvial plains of intertwined, wide and shallow river systems. Bone fragments and other bioclasts are common in the conglomeratic beds (Fernandes 2004), including isolated coprolites, scales, bones, teeth and carapace fragments, pertaining to lepisosteiforms fishes, anurans, testudines, crocodyliforms and dinosaurs (Iori et al. The Uchoa region harbours countless outcrops of the São Jos e do Rio Preto Formation. The new taxon comes from the 'Zero Um' palaeontological site ( Fig. 1), the same locality in which peirosaurid crocodyliform dentaries and the holotype of the theropod Thanos simonattoi, among other significant fossils, were found (Marinho & Iori 2011;Iori et al. 2011;M endez et al. 2014;Delcourt & Iori 2018). The site consists of several conglomeratic outcrops in a cattle pasture area owned by Milton Roberto Palharini (Ibir a County, São Paulo State, Brazil). Fossils in this locality are found in the bedrock or in isolated, weathered blocks on the ground. In particular, MPPC 01-0040 was found in an isolated block, the origin of which was probably the 'Escadaria' site (21°01 0 01.6″ S, 49°13 0 06.9″W), an exposure of conglomeratic strata subject to constant erosion, frequently revealing new fossils.

MATERIAL AND METHOD
Virtual preparation of the material Micro-CT scanning was carried out using the GE Phoenix v|tome|x S240 scanner housed at the Centro para Documentac ßão da Biodiversidade (Departamento de Biologia, Universidade de São Paulo at Ribeirão Preto), using 1000 projections, exposure time of 0.3 s, voltage of 130 kV, and a current of 150 lA, with a voxel size of 25.8 lm. We used Avizo (v. 7.0.0; Visualization Sciences Group) to virtually and manually segment skull bones and to produce virtual casts of the endocranium, endosseous labyrinth, vessels, and nerves. All 3D models are compiled and available as separate supplementary files (Hermanson et al. 2020) along with a VAXML script that can be readily open in SPIERS (Sutton et al. 2012).

Phylogenetic analysis
We added A. uchoensis and 29 other taxa to the matrix of Ferreira et al. (2018c;modified from Ferreira et al. 2018a) to evaluate the phylogenetic position of the new taxon. Some taxa were re-scored to correct previous miscodings and to add information from lCT of 31 taxa, as well as from reinterpretations of original descriptions (see Hermanson et al. 2020). Based on these data, 18 new characters were added to the data matrix, whereas 22 other characters were added from different literature sources (see Hermanson et al. 2020). The resulting charactertaxon matrix was analysed in TNT 1.5 (Goloboff & Catalano 2016) with traditional search, using 1000 replicates, hold 20, tree bi-section and reconnection (TBR), random seed 0, and collapse of zero-length branches according to rule 1, with Pro. quenstedti, Platychelys oberndorferi, Notoemys zapatocaensis, and N. laticentralis set as the outgroup taxa. All characters were equally weighted and some treated as ordered (following Ferreira et al. 2018a). The most parsimonious trees (MPTs) retrieved in the first analysis were subjected to a second round of TBR to identify extra MPTs. Consistency (CI) and retention (RI) indexes, Bremer support, and resampling values (bootstrap and jacknife) were calculated using implemented functions in TNT and are given in Hermanson et al. Diagnosis. Amabilis can be identified as a member of the Podocnemidoidae based on the presence of a processus trochlearis pterygoidei and a complete cavum pterygoidei. Amabilis differs from all other Podocnemidoidae in possessing a medially open fenestra postotica and lacking a basioccipital-opisthotic contact (also absent only in Mogharemys blanckenhorni). It is similar in size to Y. montealtensis, but smaller than all other Bauru Group turtles; the anterolateral emargination of the skull reaches the level of the orbit, similar to Bauruemys elegans, Hamadachelys escuilliei, and podocnemidids, but differing from peiropemydids; its foramen palatinum posterius is formed only by the palatine as in Pe. mezzalirai and Podocnemis spp., whereas in Y. montealtensis, Ba. elegans, Brasilemys josai, and most Podocnemidoidae it is formed by both the pterygoid and the palatine; the foramen jugulare posterius is confluent with the fenestra postotica as in Br. josai, H. escuilliei, and Por. patagonica; it lacks the anterior foramen for the palatal branch of the carotid, which is otherwise present in H. escuilliei, Ba. elegans, peiropemydids, and most podocnemidids, and lacks a horizontal occipital shelf as in Br. josai; there is no sign of an exoccipital-quadrate contact, similar to all Podocnemidoidae (except for Br. josai); the basioccipital tubera are closer to the median line of the skull, differing from podocnemidids and peiropemydids.
Amabilis uchoensis sp. nov. Derivation of name. The name refers to Uchoa municipality, where the holotype is housed.

Description
The skull of MPPC 01-0040 is smaller than that of all other Bauru Group turtles, being 2.25 cm long, 2.06 cm wide and 0.95 cm high as preserved. It is similar in size to those of Br. josai, Y. montealtensis and some extant Podocnemis spp. (Lapparent de Broin 2000;Bonin et al. 2006;Hermanson et al. 2017). The anterior region of the cavum cranii as well as the cavum tympani on the left side are filled with rock matrix. Dorsally, on the left side, the broken lateral margins of the parietal and the posterior portion of the postorbital expose the adductor fossa (Fig. 2).
Postorbital. The left postorbital is nearly complete and contacts the frontal anteromedially, the parietal posteromedially, the jugal laterally, and forms the posterodorsal margin of the orbit (Fig. 3D), as in other podocnemidoids. It is not reduced as in Podocnemis spp., in some species of which it is not even exposed in the skull roof (Gaffney 1979;Gaffney et al. 2011). The postorbital of A. uchoensis, as in all pleurodires, possesses a ventral projection that contributes to the formation of the septum orbitotemporale, where it contacts the jugal laterally and the pterygoid ventrally, at the base of the processus trochlearis pterygoidei. Anteriorly, the postorbital also contacts the palatine ventrally and the frontal medially, with which it forms the anterior surface of the septum orbitotemporale. As in other pelomedusoids, the postorbital forms the dorsolateral part of the roof of the sulcus palatino-pterygoideus ( Fig. 3D; Gaffney et al.

2011).
Palatines. The anterior parts are not preserved, but all the posterior contacts can be identified. Apart from contacting its counterpart medially, it possesses a narrow posteromedial contact with the parietal at the base of the sulcus palatino-pterygoideus (Fig. 3D), a posterior contact with the pterygoid, a dorsal contact with the postorbital, and an anterolateral contact with the jugal (Fig. 3B, D). The posterior margin of the foramen palatinum posterius is slightly incomplete, but it is clear that this foramen is restricted to the palatine (Fig. 3B), as in some specimens of Ba. elegans (e.g. LPRP/USP 0200; G.S. Ferreira, unpub. data), and in Pe. mezzalirai, and Podocnemis spp., differing, for instance, from the peiropemydids Lapparentemys vilavilensis and Y. montealtensis (Gaffney et al. 2011;Ferreira et al. 2018b), in which the foramen is instead located along the sutural contact between the palatine and the pterygoid.
Quadrate. The ventralmost portion of both quadrates is incomplete, but the cavum tympani is entirely formed by the bone, as in other pleurodires (Gaffney 1979;Gaffney et al. 2006Gaffney et al. , 2011. The quadrate contacts the prootic anteriorly and the opisthotic posteromedially with its medial surface (Figs 2B, 3A), as in most other pleurodires, except for bothremydids, namely cearachelyines and bothremydines. In these two groups the quadrate extends medially, reaches the supraoccipital and prevents the contact between prootic and opisthotic on the surface of the otic capsule (Gaffney et al. 2006). A small contact with the putative quadratojugal can be seen dorsal to the cavum tympani in MPPC 01-0040 (Fig. 3E). The fossa precolumellaris is large, as is also the case in Pe. mezzalirai and Y. montealtensis. It is larger than in Ba. elegans and other non-podocnemidid podocnemidoids (e.g. Tong & Buffetaut 1996;Lapparent de Broin 2000;Gaffney et al. 2011). The antrum postoticum is well-developed, but not as much as in Y. montealtensis (Ferreira et al. 2018b), resembling more the condition seen in Podocnemis spp. Despite being ventrally broken, it is possible to infer that the Eustachian tube and the stapes were enclosed by bone within the incisura columellae auris, as in most pleurodires (Gaffney et al. 2006(Gaffney et al. , 2011.
The quadrate forms a posteromedial process that extends from the base of the articular process for the mandible along the ventral surface of the cranium and reaches the braincase, forming a triple contact with the basioccipital and the basisphenoid (Fig. 3B), as is characteristic of Podocnemidoidea (Gaffney et al. 2006. Besides that, as a synapomorphy for Podocnemidoidae, the quadrate of A. uchoensis laterally contributes to the formation of the cavum pterygoidei (Figs 2D, 3B). The cavum pterygoidei is otherwise enclosed by the basisphenoid posteromedially, the prootic anterodorsomedially, and the pterygoid anterolaterally (Gaffney et al. 2011). The mandibular condyle is F I G . 6 . Amabilis uchoensis gen. et sp. nov., MPPC 01-0040. Endocranial cast with associated structures virtually reconstructed in: A, dorsal; B, ventral; C, left lateral view; D-E, left labyrinth in: D, left lateral; E, dorsal view. Abbreviations: aa, anterior ampulla; al, lateral ampulla; ap, posterior ampulla; asc, anterior semicircular canal; cc, crus communis; cci, canalis caroticus internus; ccv, canalis cavernosus; cer, cerebral hemispheres; cnv, canalis nervi vidiani; cst, canalis stapedio-temporalis; lab, labyrinth; lsc, lateral semicircular canal; med, medulla oblongata; pit, pituitary cast; psc, posterior semicircular canal; V-XII, cranial nerves V-XII; VII (hyo), hyomandibular branch of the cranial nerve VII. Scale bar represents 0.5 cm. Colour online. not preserved in MPPC 01-0040, but it is possible to infer its position anterior to the basioccipital/basisphenoid suture, as in most of the Podocnemidoidae (Gaffney et al. 2011;Ferreira et al. 2018b). The posteriormost portion of the quadrate is not preserved. However, it is possible to identify the fenestra postotica, an opening that allows the passage of some cranial nerves and blood vessels, such as the stapedial artery (Gaffney 1979). The posterior opening for the jugular vein, the foramen jugulare posterius, has merged with the fenestra postotica, resulting in a mediolaterally wide opening. This morphology is identical with that of other podocnemidoids such as Br. josai and Por. patagonica, but distinct from the slit-like fenestra postotica of cearachelyine bothremydids (Gaffney et al. 2006(Gaffney et al. , 2011. Pterygoid. The pterygoid contacts its counterpart medially, the parietal dorsally, the prootic posterodorsally, the palatine anteriorly, the basisphenoid posteromedially, and the quadrate posterolaterally, as in other podocnemidoids (Gaffney et al. 2011;Ferreira et al. 2018b;Fig. 2D). As typical in the Podocnemidoidae lineage, the pterygoid of A. uchoensis has a well-developed flange on its ventral surface, which approximately parallels its lateral margin (Franc ßa & Langer 2006), broadly forms the floor of the fossa pterygoidea, and forms a complete cavum pterygoidei. As in other pleurodires, the pterygoid also forms the processus trochlearis pterygoidei laterally (Gaffney 1979). The angle between the body of the process and the skull midline is almost 90° (Fig. 3A), similar to that of Y. montealtensis, Podocnemis spp., Por. patagonica, and Pe. mezzalirai Unlike H. escuilliei, Ba. elegans, peiropemydids, and some podocnemidids (Albrecht 1976;Gaffney et al. 2011;Ferreira et al. 2018b), A. uchoensis lacks a foramen anterius canalis caroticum palatinum or a palatine canal along the dorsal surface between the pterygoid and basisphenoid (Fig. 4C), through which the palatine artery usually exits the basicranium into the cavum cranii (Gaffney 1979;Gaffney et al. 2011;Rabi et al. 2013). The absence of such structures hints at the possibility of the complete loss of this artery, as previously suggested for pleurodires by Albrecht (1976). However, all the other foramina described by Gaffney et al. (2011) as present in the cavum pterygoidei, that is, the foramen nervi abducentis, foramen posterius canalis carotici interni, foramen nervi facialis, and foramen nervi vidiani, could be identified in MPPC 01-0040 (Fig. 5A).
The foramen nervi abducentis is positioned posteromedially along the basisphenoid contact, and the foramen posterius canalis carotici interni opens at the anterior edge of the cavum pterygoidei (Fig. 5A). The cavum pterygoidei is roofed by the prootic, differing from that of erymnochelyines, in which the cavity is continuous with the canalis cavernosus, as seen in Peltocephalus dumerilianus and Erymnochelys madagascariensis (Gaffney et al. 2011). The palatal (vidian) branch of the facial nerve extends to the anterior region of the crista pterygoidea (Fig. 4), opening into a small foramen (foramen anterius canalis nervi vidiani of Rollot et al. 2018) and in another posterior foramen located laterally to the suture between pterygoid and basisphenoid. The sulcus cavernosus extends parallel to the basisphenoid contact on the dorsal surface of the pterygoid.
Supraoccipital. Parts of the dorsal and posterior regions of the supraoccipital are broken in MPPC 01-0040, preventing assessment of the degree of exposure on the skull roof. Yet, most of its contacts could be identified, including those with the parietals anterodorsally, the prootics anterolaterally, and the opisthotics posterolaterally on the dorsal surface of the otic chamber (Fig. 3A). The supraoccipital also forms the dorsal margin of the foramen magnum, where it contacts the exoccipitals posteroventrally.
Exoccipital. Each exoccipital contacts the supraoccipital anterodorsally, the opisthotic anterolaterally, and the basioccipital ventrally (Fig. 3F). As in all Podocnemidoidae, it possesses no contact with the quadrate, differing from the narrow contact of Br. josai and the characteristic wide contact of bothremydids (Gaffney et al. 2006(Gaffney et al. , 2011. The exoccipital forms the lateral margin of the foramen magnum and bears two foramina for the hypoglossal nerve (XII), as in most turtles (Gaffney 1979). As in Br. josai, H. escuilliei, and Por. patagonica, the foramen jugulare posterius of A. uchoensis is laterally confluent with the fenestra postotica, not forming a separate opening (Fig. 3F)  Basioccipital. The basioccipital contacts the basisphenoid anteriorly, the quadrates anterolaterally, and the exoccipitals dorsally (Fig. 3). It contributes to the ventral third of the occipital condyle, differing from Pe. mezzalirai, Pelomedusidae and bothremydids, in which this structure is formed solely by the exoccipitals. In occipital view, the tubera basioccipitale are closer to the skull midline (Fig. 3F) Prootic. The dorsal part of the prootic contacts the parietal anteromedially, the supraoccipital posteromedially, the opisthotic posteriorly, and the quadrate laterally (Fig. 3). Ventrally, the prootic has an extended contact with the basisphenoid medially and the pterygoid laterally, so that it is almost completely covered in ventral view (Fig. 3B), as also seen in other Podocnemidoidae (Franc ßa & Langer 2006;Gaffney et al. 2011;Ferreira et al. 2018b). The prootic is partially exposed within the cavum pterygoidei. The foramen nervi facialis is positioned centrally within the prootic and on the dorsal surface of the cavum pterygoidei. As in other pleurodires, the prootic forms the posterior margin of the foramen nervi trigemini (Gaffney 1979).
Some internal structures of the prootic could be identified. On its medial wall, posterior to the foramen nervi trigemini, there is an opening shared by the facial (VII) and acoustic (VIII) nerves, the fossa acustico-facialis (Fig. 4A). Also, both the canalis cavernosus and the canalis stapedio-temporale could be digitally reconstructed. The canalis cavernosus is bordered laterally by the quadrate and medially by the prootic, extending from the foramen cavernosum, its anterior opening into the cavum cranii, posteriorly towards the fenestra postotica. The canalis stapedio-temporale branches off the canalis cavernosus at a position approximately halfway along the latter (Fig. 4A) and is also bordered laterally by the quadrate and medially by the prootic. This canal transmits the stapedial artery from the fenestra postotica to the adductor chamber, exiting through the foramen stapediotemporale, surrounded by the prootic and the quadrate bones too. This foramen opens anterodorsally in A. uchoensis (Figs 3,  4), unlike most podocnemidoids in which it is located more dorsally (Gaffney et al. 2006(Gaffney et al. , 2011Ferreira et al. 2018c).
Basisphenoid. The ventrally exposed surface of the basisphenoid is pentagonal in shape (Figs 3, 4), as in most Podocnemidoidae (Gaffney et al. 2011), and lacks a ventrally exposed anterior process that separates the pterygoids, as in chelids and some panpelomedusoids (Gaffney 1977;Gaffney et al. 2006;Ferreira et al. 2018b). The basisphenoid contacts the pterygoids anteriorly, the quadrates posterolaterally, the basioccipital posteriorly, and the prootics anterolaterally, the latter at the dorsal region of the cavum pterygoidei. On the dorsal surface of the basisphenoid, the rostrum basisphenoidale extends anteriorly over the pterygoids (Fig. 4), similar to that of other podocnemidoids (Gaffney et al. 2011).
Posteriorly, the sella turcica is deeply excavated, and the low dorsum sellae forms a well-defined margin to the sella. Each posterolateral margin of the sella turcica is pierced by a foramen anterius canalis carotici cerebralis (Rabi et al. 2013; foramen anterius canalis carotici interni of Gaffney 1979; Fig. 4C). Right and left foramina are widely separated from the skull midline. Similar to H. escuilliei, Ba. elegans and peiropemydids, the canalis caroticus internus of A. uchoensis is a distinct canal that extends posterolaterally within the basisphenoid, differing from the condition of Podocnemis spp., in which the foramen anterius canalis carotici cerebralis is extremely short, almost reducing the course of the artery to a planar opening (Gaffney et al. 2011). The foramen nervi abducentis opens at the base of the small, anteriorly projecting processus clinoideus, and pierces the basisphenoid in a straight trajectory (Fig. 4). This foramen is small, in contrast to the large opening seen in Podocnemis spp. and Pr. caiera, and also opens laterally to the basisphenoid, within the cavum pterygoidei (Gaffney et al. 2011).
Cranial endocast. The chelonian cranial endocast has a tubular shape and can be subdivided into anterior, middle, and posterior regions (Wyneken 2001;Paulina-Carabajal et al. 2013, 2017. Because of the incomplete preservation of the dorsal part of the parietal of MPPC 01-0040, as well as of the medial portion of the frontals, and the lack of the anterior part of the skull, the anterior region of the endocast of A. uchoensis could not be digitally reconstructed, precluding the visualization of its nasal cavity and olfactory nerve canal. In contrast, part of the left cerebral hemisphere was reconstructed and mirrored to generate a bilateral model (Fig. 6A-C).
The cerebral hemispheres resemble those of some fossil podocnemidoids (e.g. Laterodorsally to the pituitary, the path for the trigeminal nerve (V) projects laterally from the cranial endocast, as in all other turtles for which there are digital reconstructions or natural endocasts (e.g. Gaffney & Zangerl 1968;Paulina-Carabajal et al. 2013, 2017Mautner et al. 2017;Ferreira et al. 2018b;Lautenschlager et al. 2018). The hyomandibular branch of the facial nerve (VII) pierces the prootic in an almost straight line (Fig. 6B), perpendicular to the cerebral endocast, to the point in which it divides into two short canals (chorda tympani), next to the incisura columellae auris in the quadrate. The palatal (or vidian) branch of the facial nerve (Fig. 6) exits the pterygoid anteriorly and opens posteriorly within the cavum pterygoidei in the prootic, while the acoustic nerve (VIII) projects laterally from the same opening as the facial nerve, the fossa acusticofacialis (Fig. 4A).
The glossopharyngeal (IX), vagus (X), and accessory (XI) nerves were reconstructed as a single digital structure (Fig. 6), given that they share the same cranial path in most turtles (Gaffney 1979). As for the canals for the hypoglossal nerve (XII), two branches exit the cavum cranii and pierce the exoccipitals posteroventrally as in most turtles (Gaffney 1979), but not in Pro. quenstedti, some bothremydids, and stereogenyines, in which the two branches share a single foramen (Gaffney et al. 2006;Ferreira et al. 2015;Lautenschlager et al. 2018).
Last, only the cerebral branch of the carotid artery is present in A. uchoensis, as in the pleurodires Chelus fimbriatus, pelomedusids, araripemydids, euraxemydids and bothremydids, in which the canal for the palatal branch is absent (Gaffney et al. 2006). The cerebral canal pierces the basisphenoid anterolaterally at a slightly acute angle and enters the braincase at the foramen posterius canalis carotici interni, within the cavum pterygoidei, as in other Podocnemidoidae (Gaffney et al. 2011;Ferreira et al. 2018b).

Phylogenetic analysis
The phylogenetic analysis (see Fig. 7 for a simplified version of the strict consensus tree; the complete version can be found in the supplementary material provided in Hermanson et al. 2020) resulted in 90 MPTs with 1480 steps each (CI = 0.251, RI = 0.732; best score hit 100 times). The present results differ from those of Ferreira et al. (2018a, b) in finding araripemydids and euraxemydids nested within Pelomedusoides, and not as pan-pelomedusoids outside Pelomedusoides; the same phylogenetic position was previously retrieved by other analyses (e.g. Meylan 1996;Gaffney et al. 2006Gaffney et al. , 2011Romano et al. 2014). In all MPTs, Amabilis uchoensis is positioned as a non-podocnemidid member of Podocnemidoidae. Podocnemidoidae is characterized by the small entrance of the antrum postoticum, a complete cavum pterygoidei, parallel craniolateral margins of vertebral scale 1 (char. 207), a short path for the cerebral carotid canal that enters the braincase anteriorly to the foramen stapedio-temporale (char. 257), and foramina nervi hypoglossi ventrally covered by an extension of the basioccipital and exoccipital (char. 265).
Some of the new characters based on the available lCT data were retrieved as synapomorphic for different pleurodiran clades in the analysis. For example, an angle between the vertical (i.e. anterior and posterior) semicircular canals of the labyrinth less than 90°(char. 248) was identified as characteristic of Pelomedusoides. According to Paulina-Carabajal et al. (2017), this feature is commonly observed in turtles with a more aquatic lifestyle. However, little research has so far been done to quantitatively determine whether aspects of the labyrinth shape of turtles can be explained by ecological preferences (but see Evers et al. 2019a).
The presence of a canal for the palatal (vidian) branch of the facial nerve (char. 262) was also recovered as synapomorphic for Pelomedusoides. The clade formed by Taphrosphyini + Bothremydini (Bothremydodda of Gaffney et al. 2006) has a small opening between the lateral semicircular canal and the vestibule (char. 250), and a ratio between the lengths of the ASC and PSC of less than 1.1 (char. 252). Yet, a groove on the posterior surface of the condylus mandibularis (char. 264) was found synapomorphic for Bothremydini and for some Taphrosphyini (see Hermanson et al. 2020). Additionally, members of the clade Peiropemydidae + Podocnemididae (Podocnemidinae of Gaffney et al. 2011) have ASCs that extend dorsoventrally higher than the PSCs (char. 251). Also, a more laterally positioned path for the carotid canal (char. 258) was found only in the analysed taphrosphyines, with the angle between the foramen anterius canalis carotici cerebralis and the foramen posterius canalis carotici interni being greater than 140°.
Although not recovered as synapomorphic for any clade in the phylogenetic analysis, some other characters also deserve attention. A medulla oblongata positioned completely below the level of the cerebral hemispheres (char. 255) and a <1.1 ratio between the ASC and PSC lengths (char. 252) were found to be homoplastic, given that they are optimized as appearing independently in bothremydids and stereogenyines (which are considered littoral or marine taxa; Gaffney et al. 2006;Ferreira et al. 2015). This last feature can also be seen in some extant and extinct marine turtles (Paulina-Carabajal et al. 2013;Neenan et al. 2017). Finally, in both Phosphatochelys tedfordi and Ummulisani rutgersensis, part of the internal carotid canal is dorsally open, forming a groove between the basisphenoid and prootic (char. 260), a condition somewhat similar to that found in modern chelonioids The entire course of the carotid artery, as well as its split into the cerebral and palatal branches, is floored by bone (Sterli & de la Fuente 2010). The split of the internal carotid into two subordinate branches, however, may be absent in some pleurodires (see also Gaffney et al. 2006): only chelids and some podocnemidoids possess a foramen anterius canalis carotici palatinum, that is, the anterior opening through which the palatine artery exits the basicranium (Rabi et al. 2013), and therefore unambiguous osteological evidence for the presence of a palatine artery. It is possible, however, that the split between the two branches takes place anteriorly to the sella turcica, as suggested by Joyce et al. (2018) for carettochelyids. Also, whereas in chelids the palatine and cerebral canals possess approximately the same diameter, podocnemidoids evolved a much larger foramen and canal for the palatine branch (Fig. 8), possibly related to the development of the cavum pterygoidei.
In pleurodires that lack osteological evidence for a palatine artery, the canal of the internal carotid takes an anteromedial path that is identical to that of the cerebral artery of pleurodires with both arterial branches, i.e. it traverses the basisphenoid and exits into the sella turcica. We also found that Araripemys barretoi, previously interpreted as having the foramen for the palatine artery (e.g. Meylan 1996;Gaffney et al. 2006), also lacks that aperture. In fact, the structure identified by Meylan (1996) and Gaffney et al. (2006) as the foramen caroticum laterale is actually the opening for the abducens nerve (VI) that pierces the basisphenoid. Thus, we propose that probably the plesiomorphic condition for pelomedusoids is the absence of the foramen, which is regained in Podocnemidoidae, and lost again in the coastal/marine stereogenyines (Fig. 8).
Last, the facial nerve (VII) path is also variable in pleurodires. None of the analysed specimens possesses a foramen pro ramo nervi vidiani, which is typical of cryptodires (Gaffney 1979;Evers & Benson 2019) and baenids (at least for Eu. cephalica; Rollot et al. 2018). Chelids lack an exclusive canal for the palatal (vidian) branch of the facial nerve, which is seen in all other pleurodires (Fig. 8). A closer examination of the inner skull anatomy of pan-pleurodires such as Pl. oberndorferi and N. laticentralis (Lapparent de Broin et al. 2007) may clarify whether this canal was lost in the chelid lineage or if its absence represents the plesiomorphic condition for the group. The canal for the vidian branch was observed in all other side-necked turtles, extending from the geniculate ganglion (where the hyomandibular branch also originates; Evers & Benson 2019) to the anterior portion of the pterygoid, close to the crista pterygoidea, opening into the foramen anterius canalis nervi vidiani, as in Eu. cephalica (Rollot et al. 2018). The hyomandibular branch, in contrast, is found in all pleurodires, but in two different conditions. It may extend laterally and then bend posteriorly (Fig. 8), as seen in chelids and non-Podocnemidoidea pelomedusoids, as well as in most groups outside Pleurodira (e.g. testudinids and carettochelyids) and Testudines (e.g. baenids, plesiochelyids, and Pro. quenstedti;Paulina-Carabajal et al. 2013;Joyce et al. 2018;Lautenschlager et al. 2018;Rollot et al. 2018;and Evers & Benson 2019). In contrast, podocnemidoideans (Bothremydidae and Podocnemidoidae) have a canal for the hyomandibular branch that extends through the prootic almost entirely perpendicular to the midline axis (Figs 6, 8).
The analysis of the nerve and carotid circulation patterns of pleurodires revealed an unappreciated variation. For example, although the posterior foramen for the entrance of the carotid artery is positioned in the pterygoid-quadrate region (Gaffney et al. 2006) in the analysed Bothremydinae, that is, Bothremydini plus Taphrosphyini, a more lateralized path is found only in the latter group (Fig. 8) Instead, the foramen can be located in the pterygoid-basisphenoid region, within the cavum pterygoidei, both in some bothremydids (e.g. cearachelyines) and in Podocnemidoidae. Of these, however, the entry for the internal carotid is situated anteriorly to the foramen stapedio-temporale only in the Podocnemidoidae lineage, being relatively shorter than in the remaining pleurodiran turtles. This highlights the importance of the inner anatomy for establishing relationships in groups for which the external morphology is not variable.
According to Miyashita (2013), although the bones (basisphenoid, prootic, pterygoid and quadrate) enclosing the foramen posterius canalis carotici interni in pleurodires vary considerably (Gaffney et al. 2006), its position in relation to other cranial landmarks is 'stable within the clade' (Miyashita 2013, p. 88). Yet, our analysis of innervation and carotid circulation in pleurodires using lCT data shows that there is a previously unappreciated variation of paths and internal foramina for the cranial nerves and arteries. The variation in both the internal structures and the bones enclosing the foramen posterius canalis carotici interni might be related to the later timing of palatoquadrate fixation to the braincase in pleurodirans, which is accomplished only when dermal and endochondral ossifications are already well-developed (Werneburg & Maier 2019). The resulting variation is of interest for systematics, as shown by the present phylogenetic analysis, and should be further explored in future works.

CONCLUSION
The new podocnemidoid turtle from the Late Cretaceous São Jos e do Rio Preto Formation increases the diversity of taxa on the stem-lineage to Podocnemididae, highlighting the importance of the Bauru Group fossils in the understanding of the origins of the group (Fig. 7B). Based on an exclusive set of characters, A. uchoensis can be attributed to the Podocnemidoidae lineage. Also, based on lCT data, new characters were proposed, previous scorings were re-interpreted, and these were included in a new phylogenetic analysis. The recovering of some of these novel characters as synapomorphic for main pleurodiran clades (e.g. Pelomedusoides, Bothremydidae) indicate that, as with other vertebrates (e.g. Coates 1999;Mennecart et al. 2017;Pierce et al. 2017), neuroanatomy reflects important phylogenetic signals also for pleurodires.
Our investigation has revealed a previously unknown variety of carotid circulation patterns in pleurodires. Among Testudines, we find a single loss of the canal for the palatal (vidian) branch of the facial nerve (VII) in chelids, as suggested by Albrecht (1976). We also mapped two independent origins of a specific foramen for the palatal branch of the carotid artery, in chelids and in the Podocnemidoidae lineage, although the origin of this foramen appears to be homoplastic across turtle lineages (Fig. 8) . Furthermore, taphrosphyine bothremydids evolved one of the most lateralized paths for the internal carotid among pleurodires, in which the canal extends virtually perpendicular to the midline axis of the skull, whereas in the majority of pleurodires (and turtles in general) the carotid enters the braincase at a much more acute angle. We hope that this contribution will serve as a basis for future surveys exploring the inner skull anatomy of a larger sample of pleurodires, broadening our understanding of the evolution of the cranial innervation and carotid systems in the group.
Acknowledgements. This research was funded by FAPESP (grants 2016/03373-0 to GH, 2014/25379-5 to GSF, and 2014/03825-3 to MCL) and NERC (grant NE/L0021612/1 to SWE). We thank Daniel Cavalari (USP) for help with lCT. We are also thankful to the following curators who made specimen access and scanning possible to one of the authors (SWE): Carl Mehling (AMNH), Mark Norell (AMNH), and Yi Honyu for setting up scans at the AMNH, as well as Jessie Maisano and Tim Rowe (NSF grant IIS-9874781) from DigiMorph, and Yann Rollot (University of Fribourg) for sharing Eubaena 3D models with us. The referees Juliana Sterli, Ignacio Maniel, Sally Thomas, as well as the editor Marcello Ruta are thanked for their insightful comments that helped to improve the quality of the manuscript. This study is part of the project 'Core-facility for conservation of scientific documentation: biological collections and high technology research in comparative morphology' (CT-INFRA 01/ 2013), funded by Financiadora de Estudos e Projetos -FINEP, Ministry of Science, Technology, Innovation and Communication, Brazilian Federal Government. TNT is made freely available through the Willi Henning Society.