Geological settings
The "Turtle Cliff Fossil Site" yielded the new material described herein and is located within the Flaming Mountains about 26 km ENE of the city of Shanshan in the Turpan Basin, Xinjiang Autonomous Province, China (Figure 1). The Flaming Mountains consist of Triassic to Paleogene sediments that were uplifted during the Neogene [21–23]. Published reports on the geology and stratigraphy of the Flaming Mountains in particular and the Turpan Basin in general are rare (e.g., [24] and references therein) and many uncertainties therefore exist regarding the absolute age of formations and their correlation with similar units in other Central Asian basins. Jurassic clastic strata in the Flaming Mountains were preliminarily divided into the Early Jurassic Sangonghe Formation, the Middle Jurassic Xishanyao, Sanjianfang, Qiketai, and Qigu Formations (the latter was recently dated in the Junggar Basin with 164.6 Ma ± 1.4 Ma, [25]), and the Late Jurassic Karaza Formation [26]. Future stratigraphic research needs to clarify whether Late Jurassic strata are indeed mostly absent in the area.
Piedmont-fluvial deposits dominate the upper parts of the Jurassic sequence [27, 28]. Red-colored sediments, especially prominent in the Qigu Formation, indicate a reduction in the monsoonal circulation in Asia resulting in a paleoclimatic change from humid to seasonally dry during the late Middle and early Late Jurassic [24, 25, 28–31]. The total thickness of the supposed Qigu Formation is about 850 m in the area of the Turtle Cliff Fossil Site [31]. The formation is rich in vertebrate fossils, dominated by dinosaurs and turtles [28]. Finds of the latter include the spectacular turtle taphocoenosis at Mesa Chelonia [28] near the lower border of the formation and the herein introduced Turtle Cliff Fossil Site near the base of the upper third of the formation.
The Turtle Cliff Fossil Site is situated geographically 1 km to the ENE and stratigraphically 500 m above the Mesa Chelonia site [28]. Since no explicit justification has been given for the correlation of the strata supposedly belonging to the Qigu Formation in the Turpan Basin, the assignment of rocks units exposed in the Flaming Mountains to this formation is not transparent [28], but our preliminary classification places both sites within the Qigu Formation. The deposits that allegedly represent the Qigu Formation in the Turpan Basin are characterized by alternating coarse and fine-grained sediments that often contain unionid freshwater bivalves, reflecting changing depositional conditions typical of river systems [24, 31]. Temporary subaerial exposure is indicated by paleosols [28].
The turtle skeletons at the Turtle Cliff Fossil Site were found on the top of a low hill in a steeply inclined (65°), fine-grained and strongly cemented sandstone layer rich in lithoclasts. Above and below the turtle-bearing sandstone horizon follows a succession of predominately red silt-and mudstones.
Material studied in this paper
Our description of Xinjiangchelys wusu n. sp. is based on a sandstone slab with at least 3 individuals (Figure 2) that were excavated during the 2011 joint field season of the University of Tübingen, Shenyang Normal University, and Jilin University, that was lead and carried out by all coauthors at the Turtle Cliff Fossil Site (see Geological Settings). The quarried fossils are currently housed at the Paleontology Museum of Liaoning (PMOL) at Shenyang Normal University, Shenyang, Liaoning but will eventually be integrated into the municipal museum of Shanshan, Xinjiang Autonomous Province that is currently under construction. All specimens have been assigned a combined PMOL-Sino-German Cooperation Project (SGP) number, which will be deposited with the specimens once the museum in Shanshan is operational. The detailed coordinates of the locality are archived at PMOL and will be disclosed to qualified researchers interested in studying the site.
Specimen PMOL-SGP A0100-1 was discovered with the carapace exposed in dorsal view in 2009 below a small cliff and was cut out of the hard sandstone ledge in a block with an ICS diamond chain rock saw in 2011. Subsequent preparation revealed that the slab contained two more individuals with PMOL-SGP A0100-2 cut in half through the long axis during excavation. The slab in total includes PMOL-SGP A0100-3: shell with carapace partially exposed, femora, skull and lower jaw; PMOL-SGP A0100-2: shell (plastron not exposed), partial neck, left foot and left hand and PMOL-SGP A0100-1: posteriorly incomplete carapace, neck, crushed skull with articulated mandible, left and right hand and incomplete left posterior limb.
The anatomy of fossil taxa was reviewed mostly based on personal observations of published material and with the help of photographs. The following fossil taxa were studied first hand: Allopleuron hoffmanni (Gray, 1831) [32] (NHMUK R42913); Chubutemys copelloi Gaffney et al., 2007 [10] (MPEF-PV1236); Dracochelys bicuspis Gaffney and Ye, 1992 [33] (IVPP V4075); Hangaiemys hoburensis Sukhanov and Narmandakh, 1974 [34] (PIN 3334-4, PIN 3334-34, PIN 3334-35, PIN 3334-36, PIN 3334-37); Heckerochelys romani Sukhanov, 2006 [35] (PIN 4561-2 and PIN 4719-34); Hoyasemys jimenezi Pérez-García et al., 2012 [36] (MCCM-LH-84); Helochelydra nopcsai Lapparent de Broin and Murelaga, 1999 [37] (IWCMS 1998.21); Judithemys sukhanovi Parham and Hutchison, 2003 [9] (TMP 87.2.1); Kallokibotion bajazidi Nopcsa, 1923 [38] (NHMUK R4921 and NHMUK R4925); Kayentachelys aprix Gaffney et al., 1987 [39] (MNA V1558, MCZ 8917); Macrobaena mongolica Tatarinov, 1959 [40] (PIN 533-4); Manchurochelys manchoukuoensis Endo and Shikama, 1942 [41] (PMOL AR00008); Meiolania platyceps Owen, 1886 [42] (NHMUK R682); Mongolemys elegans Khosatzky and Mlynarski, 1971 [43] (five uncatalogued skulls at the collections of PIN); Mongolochelys efremovi Khozatsky, 1997 [44] (PIN 552-459 and two uncatalogued skulls); Naomichelys speciosa Hay, 1908 [45] (FMNH PR 273); Niolamia argentina Ameghino 1899 [46]Notoemys laticentralis Cattoi and Freiberg, 1961 [47] (cast of MOZP 2487); Odontochelys semitestacea Li et al., 2008 [48] (IVPP V13240); Ordosemys leios Brinkman and Peng, 1993 [49] (IVPP V9534-1); Peligrochelys walshae Sterli and de la Fuente, In press [16] (MACN PV CH 2017, MACN PV CH 2017); Portlandemys mcdowelli Gaffney, 1975 [50] (NHMUK R2914, NHMUK R3163, NHMUK R3164); Proganochelys quenstedti Baur, 1887 [51] (SMNS 16980); Rhinochelys elegans Lydekker, 1889 [52] (NHMUK R27); Sandownia harrisi Meylan et al., 2000 [53] (MIWG 3480); Sinemys gamera Brinkman and Peng, 1993 [54] (IVPP V9532-11); Sinemys brevispinus Tong and Brinkman, In press [55] (IVPP V9538-1); Solnhofia parsonsi Gaffney, 1975 [56] (TM 4023); Toxochelys latiremis Cope, 1873 [57] (NHMUK R4530 and NHMUK R3902); Xinjiangchelys (Annemys) levensis Sukhanov and Narmandakh, 2006 [58] (PIN 4636-4-2, [59]); Xinjiangchelys (Annemys) latiens Sukhanov and Narmandakh, 2006 (PIN 4636-6-2, [59]); and Xinjiangchelys radiplicatoides Brinkman et al., 2013 [15] (IVPP V18104).
The following taxa were studied on the basis of photographs: Adocus lineolatus Cope, 1874 [60] (CCM 60-15); Basilochelys macrobios Tong et al., 2009 [61] (MD 8-2); Bouliachelys suteri Kear and Lee, 2006 [62] (SAM P41106); Meiolania platyceps AM F: 18671; Plesiochelys etalloni Pictet and Humbert, 1857 [63] (MH 435); Pleurosternon bullockii Owen 1842 [64] (UMZC T1041).
Osteological terminology
The cranial nomenclature presented by Gaffney [65, 66] has been highly influential, because all anatomical systems of the cranium were clearly described and illustrated in these publications and because a broad audience was thereby enabled to apply these names consistently to the skulls of fossil and recent turtles. Only in the last few years have some shortcomings become apparent, however, particularly in regards to the nomenclature of the carotid system and we herein seek to rectify this situation by providing an internally consistent nomenclatural system for this anatomical region (Figure 3).
The internal carotid artery of most turtles, like most amniotes, splits into a cerebral and a palatine (lateral) branch. Although these structures are interrelated, they can be thought of as three different vessels, which are herein terms the internal carotid artery, the cerebral artery, and the palatine artery. New insights into the cranial anatomy of basal turtles [15, 20, 67] has revealed that these three blood vessels can enter the skull through three non-homologous foramina and that they can also exit the skull through three non-homologous foramina, for a total of six non-homologous foramina. The nomenclatural system of Gaffney [65, 66] proved to be confusing, because it only provides three names for these six foramina (i.e., foramen anterior [italics added for emphasis] canalis carotici interni, foramen posterior canalis carotici interni, and foramen caroticum laterale) and because these names were defined as applying to inappropriate portions of the carotid system. For instance, the foramen anterior canalis carotici interni was defined as applying to the exit of the cerebral artery, not to the exit of the internal carotid artery, whereas the foramen posterior canalis carotici interni could either be the entry of the internal carotid artery or of the cerebral artery [65, 66]. An addition oddity of this nomenclatural system that makes it difficult for neophytes to learn that the palatine artery is situated in the "lateral canal," not the palatine canal. Incidentally, the use of the "-ior" suffix (as in anterior and posterior) is not appropriate, given that the "-ius" suffix is the proper neuter singular ending in Latin.
Sterli et al. [20] were the first to realize these deficiencies in the nomenclatural system of Gaffney [65, 66] and proposed new terms, but these new terms are not sufficient to name all six potential foramina and they break with tradition set by Gaffney [65, 66] in their grammatical construction. These inconsistencies were partially addressed recently [15] but some parts of the system still remain unnamed and the palatine artery is still defined as sitting in the lateral canal.
We herein propose a new nomenclatural system that attempts to follow the grammatical precedence set forth by Gaffney [65, 66], but that breaks tradition by providing names for all potential foramina and by renaming the lateral canal the palatine canal. This nomenclatural system consists of a total of 10 new terms (Figure 3):
Canalis caroticus internus
The bony canal that holds any portion of the internal carotid artery, absent, among others, in basal turtles and paracryptodires.
Foramen posterius canalis carotici interni (fpcci)
The posterior entry of the internal carotid artery, absent, among others, in basal turtles and paracryptodires.
Foramen anterius canalis carotici interni (facci)
The anterior exit of the internal carotid artery, only present in turtles with a fenestra caroticus.
Canalis caroticus cerebralis
The bony canal that holds any portion of the cerebral artery, present in all turtles.
Foramen posterius canalis carotici cerebralis (fpccc)
The posterior entry of the cerebral artery, not developed in turtles where the split of the internal carotid artery into the cerebral and palatine branches is covered by bone.
Foramen anterius canalis carotici cerebralis (faccc)
The anterior exit of the cerebral artery, present in all turtles, typically located near the dorsum sellae.
Canalis caroticus palatinum
The bony canal that holds any portion of the palatine artery, generally absent in turtles with an open interpterygoid vacuity.
Foramen posterius canalis carotici palatinum (fpccp)
The posterior entry of the palatine artery, generally developed in turtles with a closed interpterygoid vacuity, but not in those where the split of the internal carotid artery into the cerebral and palatine branches is covered by bone.
Foramen anterius canalis carotici palatinum (faccp)
The anterior exit of the palatine artery, generally present in turtles with a close interpterygoid vacuity.
Fenestra caroticus (fca)
A figurative bony window into the otherwise closed carotid system, which exposes the split of the internal carotid artery into the cerebral and palatine branches. The window is posteriorly defined by the foramen anterius canalis carotici interni and anteriorly defined by the foramen posterius canalis carotici cerebralis and the foramen posterius canalis carotici palatinum or the interpterygoid vacuity.
Phylogenetic analysis
A phylogenetic analysis was performed using TNT [68, 69] using a modified version of a previous character/taxon matrix [16], which in return is based on earlier studies [3, 5, 59, 70] [Additional file 1]. Part of the changes are reported in an in press paper by Rabi et al. [59] and these are repeated below for the sake of clarity. Five taxa were added to the matrix [16], including Xinjiangchelys radiplicatoides, X. junggarensis (sensu Brinkman et al. 2008 [71]), X. (Annemys) levensis, X. (Annemys) latiens, and Basilochelys macrobios. The scorings of X. radiplicatoides are primarily based on the literature [15], those of X. junggarensis (=X. latimarginalis[72]) on personal observation of IVPP material from Pingfengshan [72], those of X. (Annemys) levensis, and X. (Annemys) latiens based on personal observation of PIN material, and those of B. macrobios based on the literature [60] and photographs obtained from H. Tong. The following scorings were changed relative to the original matrix [16] (the earlier scorings are in parenthesis): Epiplastron B: Hangaiemys hoburensis: 1 (?), Sinemys lens Wiman, 1930 [73] 1 (?); Pterygoid B: H. hoburensis 1 (2), Dracochelys bicuspis 1 (2), Pleurosternon bullockii 1 (2), Kallokibotion bajazidi 1 (2), Mongolochelys efremovi 1 (2), Chubutemys copelloi 1 (2), Eileanchelys waldmani Anquetin, 2009 [74] ? (2); Carapace D: H. hoburensis 0 (?), Chengyuchelys baenoides Young and Chow, 1953 [75] (IVPP-V6507) 0 (1); Carapace E: H. hoburensis-(?); Vertebral A: Siamochelys peninsularis Tong et al., 2002 [76] ? (1); Vertebral C: S. peninsularis ? (1); Anal A: S. peninsularis ? (0), Ch. baenoides ? (0); Entoplastron B: Ch. baenoides ? (1); Mesoplastron A: S. peninsularis 2 (0); Hypoplastron A: Ch. baenoides ? (0); Xiphiplastron A-B: Ch. baenoides ? (0); Dorsal Rib A: S. peninsularis ? (2); Plastral Scute B: S. peninsularis 1 (0).
Further modifications relative to Rabi et al. in press [59] include the addition of Xinjiangchelys wusu to the matrix and changing of the following scorings: Supraoccipital A: X. (Annemys) levensis 1 (0); X. radiplicatoides ? (0); X. (Annemys) latiens ? (0); Pterygoid B: Sphenodon punctatus 0 (2), Anthodon serrarius 1 (2), Peligrochelys walshae 2 (1), Niolamia argentina 2 (?); Dentary A: X. (Annemys) levensis 0 (1); X. junggarensis ? (1).
The character Cervical Vertebrae A was omitted from the analysis because we found it difficult to replicate this character objectively and perceived a number of inconsistencies in the matrix [59]. The character Diploid Number A was also omitted following previous studies [3, 59, 77].
The following characters were treated as ordered: 7 (Nasal A), 19 (Parietal H), 27 (Squamosal C), 40 (Maxilla D), 42 (Vomer A), 50 (Quadrate B + C), 52 (Antrum Postoticum A), 59 (Pterygoid B), 81 (Opisthotic C), 82 (Opisthotic D), 89 (Stapedial Artery B), 98 (Canalis Caroticum F), 120 (Carapace A), 121 (Carapace B), 130 (Peripheral A), 133 (Costal B), 138 (Supramarginal A), 158 (Hyoplastron B), 159 (Mesoplastron A), 161 (Hyoplastron B), 176 (Abdominal A), 213 (Cleithrum A), 214 (Scapula A), 232 (Manus B), 233 (Manus C). Sphenodon punctatus, Owenetta kitchingorum, Simosaurus gaillardoti and Anthodon serrarius were designated as outgroups [16, 59]. Although, there is growing evidence for a turtle-archosaur clade among molecular studies, morphological analyses still suggest lepidosaurian or parareptilian affinities for turtles at the moment. As it turns out, however, the choice of outgroup is irrelevant, as all outgroups reveal that the presence of teeth and the lack of a complete shell should be considered primitive for turtles and that the partially shelled, toothed taxon Odontochelys semitestacea is therefore sister to all turtles. The fusion of the basicranium discussed in our paper occurs far deeper within the turtle tree and is therefore not influenced by the choice of outgroups, but rather by the arrangement of basal turtles.
Given that this analysis is focused on the phylogenetic relationships and placement of xinjiangchelyid turtles, we decided to crop taxa not pertinent to these questions (e.g., most derived baenids, most meiolaniforms) and a broad spectrum of taxa known from fragmentary material only (see Appendix A for a complete list) in order to reduce the size of the matrix [59]. The resulting matrix consists of 237 characters for a total of 84 terminal taxa. The character-taxon matrix and the tnt. file are found under [Additional files 1, and 2], respectively.
The relationships of living cryptodiran taxa were manually constrained according to recent results of molecular phylogenetic studies (following previous studies [1, 2, 59]), without assuming a priori, however, that Trionychia nests within Cryptodira [78, 79]. The internal relationships of durocryptodires were constrained using a molecular topology [79] (i.e., (Emydidae (Geoemydidae + Testudinidae)) + (Chelonioidea (Chelydridae + Kinosternoidea))). The complete list of taxa designated as floaters can be found in Appendix B. A first run of heuristic search tree-bisection-reconnection, using thousands of random addition sequence replicates and 10 trees saved per replicate, failed to find all the most parsimonious trees (MPT) and therefore the heuristic search was repeated until the MPTs were found 30 times during each replicate (using the command "xmult = hits 30;"). The trees retained in the memory were exposed to a second round of tree-bisection-reconnection.
Systematic paleontology
TESTUDINATA Klein, 1760 [80]
TESTUDINES Batsch, 1788 [81]
XINJIANGCHELYIDAE Nessov in Kaznyshkin et al., 1990 [7] (sensu Rabi et al., In press [59])
Remark
We follow the phylogenetic definition of Xinjiangchelyidae used in Rabi et al. (In press [59]) where Xinjiangchelyidae is defined as the most inclusive clade containing Xinjiangchelys junggarensis Ye, 1986 [82], but not Sinemys lens, Macrobaena mongolica, or any species of Recent turtle.
Xinjiangchelys Ye, 1986 [82]
Remark: A number of genera other than Xinjiangchelys have been referred to Xinjiangchelyidae in recent years, including Chengyuchelys Young and Chow, 1953 [75]; Tienfuchelys Young and Chow, 1953 [75]; Annemys Sukhanov and Narmandakh, 2006 [58]; Shartegemys, Sukhanov and Narmandakh, 2006 [58]; Yanduchelys Peng et al. 2005 [83]; Protoxinjiangchelys Tong et al. 2012 [13] ([8, 13–15, 58, 71]). The majority of these genera are sufficiently diagnosed relative to Xinjiangchelys, but there is no up-dated diagnosis available for Xinjiangchelys. This taxon has therefore been rendered a waste-backed taxon defined by what it is not. To avoid further complications we suggest using a more inclusive definition for Xinjiangchelys that includes all species of Xinjiangchelyidae (sensu Rabi et al., In press [59]) until the phylogenetic relationships of the included taxa can be determined more confidently.
Xinjiangchelys wusu sp. nov.
(Figure 2, Figures 4, 5, 6, 7, and 8)urn:lsid:zoobank.org:pub:2BCCC095-7622-4 F27-8 F80-6199F24690B5
Holotype
PMOL-SGP A0100-1, a partial skeleton, including the skull exposed in dorsal view (Figures 2, 4, 6 and 8B).
Referred material
PMOL-SGP A0100-3, partial skeleton (Figures 2, 4, 5 and 7,); PMOL-SGP A0100-2, partial skeleton without skull, plastron not exposed (Figures 2, 6 and 7A, C-D).
Locality and horizon
Turtle Cliff Fossil Locality (see Geological Settings), Shanshan, Turpan Basin, Xinjiang Autonomous Province, People’s Republic of China (Figure 1); ?Qigu Formation, Upper Jurassic.
Etymology
wusu refers to a small town in Xinjiang Autonomous Province.
Diagnosis
A species of Xinjiangchelys; skull differing from X. (Annemys) levensis in the prefrontals being fully separated by the frontals; from X. (Annemys) latiens by the broader skull and the extensive jugal and frontal contribution to the orbit, from X. radiplicatoides by the flattened skull and the presence of a remnant of the interpterygoid vacuity. Shell differing from X. chowi Matzke et al. [84]X. qiguensis Matzke et al., [85]X. tianshanensis Kaznyshkin et al. [7] and X. junggarensis (sensu Brinkman et al. 2008 [70]) by the narrow vertebral scales.
Description
Skull
Preservation
The skull of PMOL-SGP A0100-1 is exposed only in dorsal and lateral views, whereas its palatal side is covered by the carapace of PMOL-SGP A0100-2 (Figure 2). It is dorsoventrally crushed and the preorbital region is slightly shifted from its original position. The skull of PMOL-SGP A0100-3, on the other hand, is exposed in ventral view and in articulation with the hyoids and the mandible (Figures 4B and 5).
Scales
Some of the cranial scales are traceable in PMOL-SGP A0100-1, but most of them are not apparent (Figure 4A). Using a recently suggested nomenclatural system [16] we identify the unpaired scale Y on the posterior half of the frontal posteriorly bordered by the paired scale F. Scale G is bordered by the unpaired parietal scale X posteriorly. Scale A is another unpaired scale of the parietal found posteriorly to scale X. Scale H may have also been present laterally to scale X. The skull roof is otherwise decorated with fine grooves and very shallow pits that do not show a clear pattern.
Nasals
The nasals are very poorly preserved but their sutures with the frontal and the prefrontal are partially traceable on the right side of PMOL-SGP A0100-1 (Figure 4A). They are reduced, posteriorly tapering elements that are partially separated by the anterior frontal process. The nasals contribute to the formation of the external nares.
Prefrontals The dorsal plate of the prefrontals is elongate and medially separated from its counterpart by the anterior frontal process (Figure 4A). The prefrontal contacts the nasal anteriorly and the maxilla ventrally. The descending process of the prefrontal has a wide contact with the palatines within the fossa orbitalis. Its contact with the vomer is not visible, but given the large sirms the anterior half of the dorsal margin of the orbit.
Frontals The frontals form an anterior process that is wedged between the prefrontals (Figure 4A). The posterior half of the dorsal margin of the orbit is formed by the frontals. The orbit has a subcircular outline and faces dorsolaterally.
Parietals
The dorsal plate of the parietals exhibits a relatively deep temporal emargination that reaches beyond the level of the anterior border of the cavum tympani (Figure 4A). The parietal meets the frontal anteriorly and has a long contact with the postorbital. Even though the parietals are slightly shifted from their original position in PMOL-SGP A0100-1, their posterolateral tips also touched the squamosal, as seen on the right side. Dorsoventral crushing obscures the structures of the processus inferior parietalis.
Jugal
The jugal area is compressed and its entire lateral surface is exposed in dorsal view in PMOL-SGP A0100-1 and it is also partially visible in PMOL-SGP A0100-3 (Figures 4A and 5B). It sends a long posterior process along the postorbital but it is unclear whether it meets the quadratojugal. The skull exhibits a moderate cheek emargination that exposes the coronoid process of the mandible. Anteriorly, the jugal forms the posterolateral margin of the orbit and contacts the maxilla. It is unclear whether the ventral plate of the jugal contacts the posterior end of the triturating surface and/or the anterolateral tip of the external pterygoid process.
Quadratojugal The quadratojugal is a reduced, flat element that is best preserved in PMOL-SGP A0100-3, though not fully exposed (Figure 5B). It has no clear contact with the jugal anteriorly but this is all but certain. Dorsally, it meets the postorbital and posteriorly it borders the cavum tympani. Its lower rim of the skull is emarginated, which gives the quadratojugal a subtriangular outline. The quadratojugal sends a pair of narrow and tapering processes along the dorsal and the ventral margins of the cavum tympani, respectively. The dorsal one of these processes is wedged between the postorbital and the quadrate and appears not to reach the squamosal. The ventral one terminates slightly before the level of the condylus mandibularis.
Squamosal
The squamosal is better preserved on the right side of PMOL-SGP A0100-1, the left one being compressed and the lateral plate being exposed when the skull is viewed dorsally (Figure 4). The lateral surface of the squamosal is smooth and there is no squamosal horn. The squamosal has a very short point-like contact with the parietal along the anterior margin of the upper temporal emargination. There is no contribution to the formation of the anterior opening of the antrum postoticum as seen in PMOL-SGP A0100-3. Medially, the squamosal contacts the quadrate and may even have a short contact with the opisthotic within the upper temporal fossa. The contact of the ventral portion with the opisthotic and the quadrate is not exposed in either specimen.
Postorbital
he postorbitals are long elements; they form the posterodorsal margin of the orbit and also contribute to the rim of the upper temporal emargination (Figure 4A). The postorbital has an anteroventral contact with the jugal, a posteroventral contact with the quadratojugal, a long lateral contact with the parietal, and also meets the squamosal posteriorly.
Premaxilla The premaxillary region is shifted anteriorly from the original position and damaged in PMOL-SGP A0100-1 (Figure 4). The premaxilla forms the ventral margin of the external nares and contacts the other premaxilla medially and the maxilla posterolaterally. The external nares are undivided. Only little of the ventral aspect of the premaxillary region is exposed but it is apparent that there is no premaxillary hook.
Maxilla The maxilla forms the ventral margin of the orbit, sends a dorsal process to contact the descending pillar of the prefrontal, and contacts the premaxilla anteriorly and the jugal posteriorly (Figures 4 and 5B). The triturating surface is only partially exposed but it is apparently narrow and straight with a sharp and low labial ridge.
Vomer
A single, slightly damaged and displaced, elegant vomer is present in PMOL-SGP A0100-3 exposed in dorsal view (Figure 4B). Its outline is very similar to that of Xinjiangchelys levensis.
Palatine The right palatine is preserved incompletely and shifted from the original position in PMOL-SGP A0100-3 (Figure 4B). It shows an extensive free lateral margin that is indicative of a large foramen palatinum posterius.
Quadrate
Apart from the region of the cavum tympani, the right quadrate of PMOL-SGP A0100-1 is in good condition whereas the left otic region is badly fragmented and compressed (Figures 4 and 5B). In PMOL-SGP A0100-3 the region of the cavum tympani is exposed in lateral view. The cavum tympani is anteriorly bordered by the quadratojugal and by the squamosal dorsally and posteriorly. The incisura columella auris is an open but tight notch and there is no precolumellar fossa. The antrum postoticum is well developed and its opening is formed entirely by the quadrate, although the squamosal comes very close to the lateral rim. The quadrate contacts within the upper temporal fossa the squamosal posterolaterally and the opisthotic medially, but its medial contact with the prootic is obscured. Together with the prootic it forms a large foramen stapedio-temporale. The quadrate forms a poorly developed processus trochlearis oticum that is composed of a rugose area.
Epipterygoid The epipterygoids are not exposed in either specimen.
Pterygoid The pterygoids are almost intact in PMOL-SGP A0100-3 except for their anteriormost edges (Figures 4B and 5A-B). The pterygoid has a long posterior process reaching as far as the back of the skull and terminating slightly anterior to the basioccipital-basisphenoid suture. The pterygoid covers the cranioquadrate space and contacts the posterolateral corner of the basisphenoid but not the basioccipital. The pterygoid has a short dorsal contact with the exoccipital, but this contact is not part of the skull surface. The foramen posterius canalis carotici interni opens at the back of the skull within the ventral surface of the pterygoid. The quadrate ramus of the pterygoid bears a well-developed, oval-shaped pterygoid fossa. The processus pterygoideus externus is present and it is characterized by a posteriorly extending horizontal plate and a dorsoventrally thickened vertical plate. A characteristic feature of the pterygoid is a large oval opening just anterior to the basisphenoid and posterior to the region where the pterygoids meet one another along the midline. This opening has intact margins, is clearly not a result of erosion or any other taphonomic processes, but is distinct from the foramen posterius canalis carotici palatinum. We interpret this structure as the remnant of the interpterygoid vacuity. Anterolaterally, the pterygoid bears a margin that is indicative of a large foramen palatinum posterius.
Supraoccipital Much of the crista supraoccipitalis is displaced and preserved in fragments in PMOL-SGP A0100-1 (Figures 4 and 5A). The supraoccipital provides only a small contribution to the skull roof where it contacts the parietals. The ventral plate of the supraoccipital contacts the opisthotic laterally and forms the dorsal margin of the foramen magnum. The crista supraoccipitalis extended apparently only slightly beyond the posterior tip of the squamosals. In PMOL-SGP A0100-3 the supraoccipital crest is intact as exposed in ventral view and does not protrude much beyond the level of the occipital condyle.
Exoccipitals The exoccipitals are preserved on both sides in PMOL-SGP A0100-3 (Figures 4 and 5A). They form the ventrolateral wall of the foramen magnum. A pair of foramen nervi hypoglossi pierce each elements but the formed foramen jugulare posterius is not distinct from the fenestra postotica (Figure 5A). Laterally, the exoccipitals contact the opisthotic and have a ventromedial contact with the basioccipital. Anteroventrally, the exoccipital has a short contact with the posteriormost tip of the pterygoid, but this contact does not contribute to the smooth, palatal surface of the skull. A suboval, unossified area excludes the exoccipital from anteromedially contacting the basisphenoid.
Basioccipital The basioccipital has a pair of basioccipital tubera with rounded posterior edges that extends as a roof over the foramen nervi hypoglossi when the skull is viewed ventrally (Figures 4B and 5A-B). The neck of the basioccipital condyle is short and lacks paired ridges or grooves. The basioccipital has no contact with the pterygoid and the processus interfenestralis of the opisthotic is therefore visible in ventral view. Anteriorly, the basioccipital meets the basisphenoid via a transverse suture. A shallow concavity extends on the ventral surface of the basioccipital that barely protrudes onto the basisphenoid.
Prootic The prootic is barely exposed on the right side of the skull of PMOL-SGP A0100-1 (not visible in dorsal view due to crushing). The skull roof in this specimen is deformed and thereby obscures the dorsomedial third of the ear capsule. On the left side, the ear capsule is so crushed that the structures cannot be identified with confidence. The prootic contributes to the large foramen stapedio-temporale together with the quadrate and maybe even with the opisthotic. There seem to be no prootic contribution to the processus trochlearis oticum (i.e. the rugose surface on the anterodorsal portion of the quadrate) by the prootic.
Opisthotic The opisthotic is exposed in both skulls (Figures 4 and 5A-B). The supraoccipital has a thin lateral lamina that partially covers the opisthotic within the upper temporal fossa. The opisthotic has a long lateral contact with the quadrate, may have a short contact with the squamosal, and ventrolaterally contacts the exoccipital. The dorsal portion of the opisthotic has a sutured contact with the quadrate whereas its contact with the squamosal is covered by matrix. The opisthotic forms a pillar-like processus interfenestralis that is visible in ventral view.
Basisphenoid The basisphenoid is preserved in good condition in PMOL-SGP A0100-3 (Figures 4B and 5B). It contacts the basioccipital along a straight suture posteriorly and is surrounded by the pterygoid rami laterally. The ventral surface is smooth and paired pits are therefore absent. The basisphenoid has a marked basipterygoid process in a form of a triangular, flat, horizontal plate that is sutured to and fits into a slightly-raised "pocket" of the pterygoid (Figure 4B). The foramen posterius canalis carotici interni is limited to the pterygoid but at least the anterior half of the floored canalis carotici interni extends along the pterygoid-basisphenoid suture. More anteriorly the carotid artery was exposed in a relatively deep and short, anteromedially directed sulcus, the fenestra caroticus, in the basisphenoid in which the split of the cerebral and palatine arteries was located. The cerebral branch diverged anteromedially and reentered the skull via the foramen posterius canalis carotici cerebralis at the medialmost corner of the fenestra. After exiting the fenestra, the palatine branch extended anteriorly in a shallow groove and entered the skull via the foramen posterius canalis carotici palatinum, which is situated on the basisphenoid-pterygoid contact just lateral to the residual interpterygoid vacuity. Since the latter foramen is clearly present, we infer that the palatine branch entered the skull here and not via the interpterygoid vacuity. The reduced condition of the interpterygoid vacuity in PMOL-SGP A0100-3 could represent a transitional state between a fully formed interpterygoid vacuity as seen in basal turtles [19, 35, 86, 87] and a completely closed one as seen in numerous crown-group turtles [3].
Mandible
The elegant and shallow mandible is preserved in articulation in PMOL-SGP A0100-3 (Figure 4B and 5B-C) and PMOL-SGP A0100-1 (Figure 4A), the former exposing the left coronoid region and the lateral plate of the dentary whereas the latter exposing the entire ventral and lateral aspects.
The dentary is characterized by a narrow triturating surface and a fused symphysis but neither the dentary ridges nor the anterodorsal tip of the symphyseal region are exposed. Laterally, the dentary extends posteriorly to meet the angular and the surangular whereas its contact with the articular is uncertain.
The coronoid is rather low and a long. An anteriorly tapering splenial is present that extends below the Meckelian canal along the dentary and approaches the symphysis. The splenial sends a posterior process between the angular and the prearticular, whereas the angular sends a similarly long anterior process into the splenial ventral to this projection. At the anteroventral tip of the angular process there is a triple junction with the dentary and the splenial. The processus retroarticularis is short. The splenial has a short dorsal contact with the coronoid.
Hyoid apparatus
Both cornu branchiale I are preserved almost in situ in PMOL-SGP A0100-3 (Figures 4B and 5A-B), the right one being slightly crushed and incomplete. There is no evidence of an ossified corpus hyoidis or cornu branchiale II and these structures were therefore likely cartilaginous. The cornu branchiale I is a single, elegant element that can be divided into an anterior horizontal half and a posterior vertical half. It tapers posteriorly and terminates in a narrow, whip-like structure. When the skull is viewed from laterally, the border of the vertical and the horizontal portion is roughly at the level of the posterior rim of the cavum tympani.
Shell
Carapace The carapace is present in all three specimens (Figures 3, 6 and 7A). PMOL-SGP A0100-1 and PMOL-SGP A0100-2 cover the anterior third of the carapace of PMOL-SGP A0100-3. PMOL-SGP A0100-1 has a posteriorly incomplete and slightly anterodorsally compressed carapace whereas PMOL-SGP A0100-2 is considerably deformed along its long axis and its right lateral third is missing due to damage that occurred during recovery of the block (see Materials and Methods). PMOL-SGP A0100-3 is not deformed; the exposed portion preserves the original outline of the carapace suggesting a relatively wide shell.
Carapacial bones The nuchal is a trapezoidal element and more than twice as wide than long (Figure 6). The nuchal emargination is minor in PMOL-SGP A0100-1 but appears to be slightly deeper in PMOL-SGP A0100-2. This emargination extends onto peripheral 1 in both specimens.
There are eight pairs of costal bones, all of which have firm contacts with the peripherals and lack costal fontanelles. The reduction of neural 7 in PMOL-SGP A0100-2 allows for a short, medial contact of costals 7, which contrasts the morphology of PMOL-SGP A0100-3, where a subdivided neural 7 does not allow for this contact (Figures 6B and 7A). This region is not preserved in PMOL-SGP A0100-1.
Costal 1 tapers laterally and is subequal in anteroposterior length with the more posterior costals. The figure of PMOL-SGP A0100-2 (Figure 6B) hints at a seemingly longer costal 1, but this is an optical illusion resulting from distortion. Costal 2 has a slightly concave anterolateral outline and sends a wide rectangular posterolateral process into the posterior half of peripheral 4, as is best seen on the left side of PMOL-SGP A0100-1 (Figure 6A), but also visible on the right side and in PMOL-SGP A0100-2 (Figure 6B). Costal 3 is the mediolaterally widest element and has straight and parallel anterior and posterior sides. Costal 4 is slightly concave anteriorly and convex posteriorly and has strongly concave contacts with peripheral 6 and 7. Costal 5 is slightly convex anteriorly and posteriorly. It has a short, oblique contact with peripheral 7 and a strongly concave contact with peripheral 8. PMOL-SGP A0100-3 is different in that costal 5 barely touches peripheral 7 (Figure 7A). The contact of costal 6 with peripheral 8 projects more laterally relative to its contact with peripheral 9. Costal 7 has an oblique and straight anterior border and a concave posterior border. Costal 8 is narrow and slightly convex anteriorly and posteriorly.
The neural series is complete and consists of eight elements, including a subdivided neural 7 in PMOL-SGP A0100-3 (Figure 7A). In PMOL-SGP A0100-2 the series is interrupted by the short contact of costals 7 (Figure 6B). In PMOL-SGP A0100-1, this region is incomplete. Most neurals are hexagonal, coffin-shaped elements with the short sides facing anterolaterally. Neural 1 of PMOL-SGP A0100-1 and 3 is quadrangular whereas it is hexagonal with short sides facing posterolaterally in PMOL-SGP A0100-2 the quadrangular element being neural 2 instead. In PMOL-SGP A0100-3 neural 7 is subdivided into a larger, regular, hexagonal element and a small, square element and neural 8 is hexagonal. In PMOL-SGP A0100-2 both neurals 7 and 8 are pentagonal and do not contact one another, thereby allowing for a medial contact of costals 7.
There are two suprapygals, the anterior one is trapezoidal has no contacts with the peripherals and considerably wider and shorter in PMOL-SGP A0100-2 than in PMOL-SGP A0100-3 (Figures 6B and 7A). Suprapygal 2 is a wide element that contacts costal 8 and peripheral 11.
There are 11 pairs of peripherals (Figures 6 and 7). A distinct gutter extends from the lateral corner of the nuchal to peripheral 7 along the lateral margin of the carapace. Peripheral 1 contacts costal 1 and is larger on the right side than on the left in PMOL-SGP A0100-1 (Figure 6A). Peripheral 2-6 are narrow elements whereas 7-11 are considerably expanded laterally. Peripheral 8 is the widest peripheral element and has a strong medial projection into costal 5 in all specimens.
Carapacial scales There are five vertebrals and a single wide cervical (Figures 6 and 7B). Vertebral 1 is wider than long and barely touches peripheral 1. The proportions of vertebrals 2-4 vary somewhat among the specimens, but all are narrower than most pleurals. PMOL-SGP A0100-1 has the relatively widest vertebrals of all. Vertebral 2 is slightly longer than wide in PMOL-SGP A0100-3 (as reconstructed) and PMOL-SGP A0100-2 whereas in PMOL-SGP A0100-1 it is markedly wider than long. Vertebral 3 is slightly wider than long in PMOL-SGP A0100-3 and PMOL-SGP A0100-3-3 and slightly longer than wide in PMOL-SGP A0100-2. Vertebral 4 is considerably wider than long in PMOL-SGP A0100-3 and PMOL-SGP A0100-3-3 and this is less distinct in PMOL-SGP A0100-2. The vertebral 3-4 sulcus has anterior projection at the midline that extends onto the posterior portion of neural 5 in all specimens. Vertebral 5 is wider than long, contacts peripherals 11 laterally, and does not prolong onto the suprapygal (as preserved in PMOL-SGP A0100-1 and PMOL-SGP A0100-3).
The pleurals are all wider than long except for pleural 4 that is longer than wide (Figures 6 and 7A).
The marginals are either restricted to the peripherals or their borders coincide with the costo-peripheral contacts or, as in the case of marginals 5 and 7, they slightly lap onto the costals. Marginal 11 indistinctly prolongs onto the costals on the left sides of all specimens (right side not preserved in PMOL-SGP A0100-1).
Plastron The plastron is only exposed in PMOL-SGP A0100-3 (Figure 7B). In this specimen the plastron is preserved in perfect condition except for minor damage in the right bridge area. The dorsal aspect of the plastron is not visible. The plastron is characterized by complete ossification (i.e., no fontanelles) and compact, sutural contacts. Scale sulci are clearly developed. The anterior lobe is about 40% wider than long, shorter than the posterior lobe, and has a slightly rounded anterior margin. Mesoplastra are absent. The posterior lobe is posteriorly tapering, slightly wider at its base than long, and lacks an anal notch. At least one musk duct foramen is present between peripheral 4 and the hyoplastron.
Plastral bones The epiplastron is trapezoidal, shows a roughly transverse suture with the hyoplastron, an anteromedially directed contact with the entoplastron, and a sagittal contact with the other epiplastron (Figure 7B). The entoplastron is oval-shaped, about twice as long as wide, and only partially separates the epiplastra. The buttress of the hyoplastron is relatively low and it terminates on the anterior half of peripheral 2. The contact of the plastron with the carapace is tight but we interpret it as being ligamentous rather than sutured owing for the presence of plastral pegs. However, we note that our meaning of ligamentous contact is probably different from the concept of earlier studies [13, 14]. The edges of the bridge peripherals slightly overlap the margin of the bridge of the plastron. Peripherals 3, 4, and the anterior third of peripheral 5 contact the plastron via well-developed pegs. Further posteriorly, the contact between the plastron and the carapace transfers into a smooth-edged contact until the posterior third of peripheral 6. Just medial to this edge the plastron is notched at the contact of the hyo-and the hypoplastron, but this space is filled up with two elements (on one side) that appear to be aberrant extra ossifications that meet the hyo- and the hypoplastron along finely serrated edges. The anterior edge of these elements is partially fused with the hyoplastron. More posteriorly, the hypoplastron contacts the peripherals via pegs with the inguinal buttress terminating on the anterior third of peripheral 8. The xiphiplastra are well developed and they have a fork-like contact with the posterolateral portion of the hypoplastron in ventral view. Interfingering interplastral sutures are absent.
Plastral scales One pair of gulars and one pair of extragulars are present. The gulars do not extend onto the entoplastron and the extragulars have a transverse contact with the humeral scales. The midline sulcus of the plastron is straight instead of sinusoidal. The pectoral scale is shorter than the abdominal. The femoral/anal sulcus is omega-shaped and the anals barely extend near the midline onto the hypoplastron. Four pairs of inframarginals are present, of which the third covers the hyo/hypoplastral suture (Figure 7B).
Appendicular skeleton
The articulated distal half of the right fore limb of PMOL-SGP A0100-1 is exposed in antipalmar view next to the skull of PMOL-SGP A0100-3 (Figures 7B and 8B). The left distal fore limb of this individual is less complete and preserved tucked on the other side of the slab next to the carapace (Figure 6A). The articulated left fore and hind limbs of PMOL-SGP A0100-2 (Figure 6B) are preserved in palmar and antipalmar view, respectively (Figures 8C-D). Only the left ulna, radius, an associated phalanx, the distal end of the right humerus, and both femora are exposed in specimen PMOL-SGP A0100-3 (Figure 7B). An additional, isolated left hind limb is present on the slab that likely belongs to a fourth specimen (Figure 3).
Humerus The humerus has a slightly curved shaft with a suboval cross-section (Figure 8C). The lateral process is slightly better developed than the medial process and both processes are situated at the same level relative to one and another along the proximal part of the humerus. The ectepicondylar foramen is closed.
Radius and ulna The radius is elegant and narrow and has a straight and relatively flat shaft in cross-section (Figures 8B-C). Its proximal epiphysis is subcircular in cross-section whereas its distal epiphysis is expanded and more compressed. The articulation surface for the medial centrale extends along the distal margin of the epiphysis as in Podocnemis expansa[19]. The medial edge of the distal epiphysis lacks a medial projection that is otherwise present in Macrochelys temminckii. The lateral ridge below the proximal epiphysis, presumably for the attachment of the radio-ulnar ligament [19], is reduced. The ulna is flattened and more robust than the radius. The medial margin of the shaft is more curved than the lateral one. The ridge for the bicipital tendon attachment is reduced and the olecranon is poorly developed. The medial process of the proximal epiphysis is situated slightly below the level of the olecranon as in M. temminckii but unlike in P. expansa. The relative proportions of the proximal and distal epiphyses more resemble M. temminckii in having similar width (the distal being slightly wider).
Manus The relatively elongate and narrow phalanges of the manus suggest intermediate aquatic adaptation [88]. The phalangeal formula is 2-3-3-3-3 (Figures 8B-C). The unguals are clawed, narrow, and pointed, and decrease in size from the digit I to V. The distal articulation surfaces of the proximal phalanges exhibit posteriorly projecting flanges that underlap the proximal epiphysis of the preceding metacarpals. The first metacarpal is the shortest and the most robust. The lateral overlapping of the metacarpals with one another is present but not marked. The distal carpals are ovoid and that of the first digit is slightly wider than those of the remaining digits. There is a small pisiform and the medial centrale is tightly connected with the lateral centrale. The intermedium is not elongate proximodistally, the ulnare is flat and deep, and the radiale bears little if any articulation with the radius.
Femur The femur has a slightly curved shaft (Figures 7B and 8D). The trochanter minor faces anteriorly, the trochanter major faces dorsally, and the femoral head only slightly extends above the trochanters. The trochanters are moderately developed. The proximal epiphysis has a similar width as the distal one.
Tibia and fibula The tibia has a wide proximal epiphysis (Figure 8D) than Podocnemis expansa or Macrochelys temminckii. The ridge for the patellar tendon attachment is placed close to the midline of the shaft as in M. temminckii and unlike in P. expansa where it is shifted laterally. The fibula is straight and has a more expanded and more compressed distal epiphysis than its proximal one. Proximally, the shaft lacks a medial flange, unlike in Podocnemis expansa.
Pes The hooked fifth metatarsal is a large, blocky element (Figure 8D). The astragalus is fused with the calcaneum. The pedal formula is 2-3-3-3-? and digits 1-4 were clawed, whereas digit 5 is incompletely preserved. The first metatarsal is more robust than the others.
Vertebral column
Four cervicals are preserved in PMOL-SGP A0100-2, three in PMOL-SGP A0100-1, and PMOL-SGP A0100-3 exhibits one cervical vertebra and two anterior caudals (Figures 4A, 6 and 8A).
In PMOL-SGP A0100-2 three cervicals are well exposed in lateral and dorsal views that could represent any series of cervicals between 2 to 6 (Figure 8A). In PMOL-SGP A0100-1 cervicals 2 and 3 are exposed in dorsal view (Figure 4A). The centra are amphicoelous and more than twice as long than high (excluding the ventral keel and including the dorsal spine). A low ventral keel extends along the entire midline of the centra. The transverse processes are compressed, relatively robust, with parallel anterior and posterior sides, and exhibit clear diapophyses. The transverse process does not extend much laterally and is slightly longer than wide. The posterior third of the transverse process extends beyond the middle of the centrum whereas its anterior two-thirds extend anteriorly to the middle of the centrum, terminating well before the anterior end of the centrum. The cervicals have well-developed bifurcated ribs. The dorsal articulation of the ribs with the transverse processes is not preserved and elements therefore must have shifted, but the anterior contact with the parapophysis is still preserved. The parapophyses are situated at the anteroventral margin of the centrum and is best developed in the second cervical preserved in PMOL-SGP A0100-2 (probably cervical 3 or 4). The neural arch is longer and more than twice as high as the centrum (centrum including the transverse process but excluding the ventral keel and the arch including the zygapophyses but excluding the dorsal spine). The postzygapophyses are only slightly separated and unite in a common low stem. The anterodorsal surface of the postzygapophysis is convex whereas the posterodorsal is concave with a groove extending anteromedially. The neural spines are damaged and their full height is therefore unknown, except for the most anterior preserved cervical in PMOL-SGP A0100-2. Cervical 2 has a long neural spine extending all along the dorsal surface of the arch whereas cervical 3 has a shorter spine (PMOL-SGP A0100-1). The anteriormost cervical in PMOL-SGP A0100-2 has a low but long spine, the following is higher, and the third has a short and high process. The prezygapophyses are a little higher than the postzygapophyses (except for the third preserved in PMOL-SGP A0100-2) and slightly extend beyond the level of the anterior edge of the centrum in lateral view.