Tacuarembemys kusterae, gen. et sp. nov., a new Late Jurassic–?earliest Cretaceous continental turtle from western Gondwana

ABSTRACT A new continental turtle, Tacuarembemys kusterae, gen. et sp. nov., is described on the basis of a partial external mold of the carapace and associated shell bone fragments recovered from the Batoví Member (Late Jurassic–?earliest Cretaceous) of the Tacuarembó Formation, Paraná Basin, Uruguay. The estimated length of the carapace is 18 cm. This new genus and species shows a unique combination of characters: a large nuchal notch, a pair of anterior supernumerary scales, the absence of a cervical scale, and an external surface ornamentation that is macroscopically smooth with some thin linear ridges perpendicular to the margins of the plates and microscopically composed of small, randomly distributed pits. The first two characters resemble those seen in the solemydid Naomichelys speciosa from the Cretaceous of North America, although the ornamentation is markedly different. Although this combination of characters—some shared with other taxa (including cryptodires and pleurodires) and some others that are autapomorphic—allows the recognition of a new genus and species, additional remains are yet needed in order to clarify its phylogenetic relationships. Tacuarembemys kusterae is part of the Priohybodus arambourgi Assemblage Zone, which is of Late Jurassic–?earliest Cretaceous age. This is the first turtle to be discovered in South American continental deposits of that age and thus increases the knowledge on the regional evolution of Mesozoic turtles. The paleoenvironment for this species includes lakes and permanent and ephemeral rivers in arid-to-semiarid climatic conditions. SUPPLEMENTAL DATA—Supplemental materials are available for this article for free at www.tandfonline.com/UJVP


INTRODUCTION
Although turtles from Upper Jurassic-Neocomian continental deposits are common in Asia, Europe, and North America (e.g., Gaffney, 1979;Hirayama et al., 2000;Joyce, 2000;Sukhanov, 2000;Milner, 2004;Rabi et al., 2010;Joyce et al., 2011;P erez-Garc ıa and Ortega, 2011), they are virtually unknown in Gondwana. In South America, only the Early-Middle Jurassic Condorchelys antiqua Sterli, 2008, and an Aptian-Albian turtle fauna from Cerro Barcino, Argentina, and the Araripe Basin of Brazil are known (Gaffney et al., 2006(Gaffney et al., , 2007de la Fuente et al., 2011;. Here we describe a new genus and species of turtle from the Tacuaremb o Formation (Late Jurassic-earliest Cretaceous), which represents the oldest known turtle from Uruguay. Furthermore, this discovery begins to fill in the fossil record gap of continental turtles from the Jurassic to the Cretaceous in South America. Although the preserved remains of this new turtle are fragmentary, we considered that its significance regarding time, environment, anatomy, and taxonomy are worth exploring. Consequently, the main goals of this paper are to present, describe, and illustrate this new turtle, to explore its phylogenetic relationships in a global turtle data set, and to discuss the importance of this finding in the global context of turtle evolution.
One of the most recent discoveries in the Batov ı Member is an external partial mold of a carapace and associated shell bone fragments, described herein, that represent a new genus and species of turtle. The remains were recovered from fine-grained, pink-yellowish sandstones of mainly quartz composition. Bioturbation is common in the vicinity of the outcrop.

MATERIALS AND METHODS
The specimen is housed in the MGT. A sandstone block containing the fossil was isolated from the rock and protected by a plaster jacket. Then, the mold was prepared under a binocular microscope and completely reinforced with cyanoacrylate. A computed tomography analysis was made so as to reject or confirm the presence of bones potentially included in the sandstone block. Standard petrographic thin-sections of shell bone samples were prepared for histological analysis. The shell surface pattern was analyzed using a JSM-5900 LV scanning electron microscope (SEM). Taxonomic nomenclature in this paper follows Joyce et al. (2004), measurements and carapace terminology mainly follow Lyson et al. (2011), and histological terminology follows Scheyer and Sander (2004) and Scheyer (2007). The stratigraphic framework follows Perea et al. (2009).
The phylogenetic analysis is based on . We added Tacuarembemys kusterae and a character (anterior supernumerary scale) to the Sterli and de la Fuente (2013) data set, resulting in an augmented data matrix of 102 taxa and 241 characters, which was constructed in Mesquite (Appendix S1; Maddison and Maddison, 2007). All characters showing a clear morphocline were considered additive (22 in total; see Appendix S2). All characters were weighted equally. All cladistic analyses were performed in TNT (Goloboff et al., 2008a(Goloboff et al., , 2008b. Two rounds of tree bisection reconnection (TBR) were performed. In the first round, the original Wagner tree is replicated and perturbed 1000 times using the TBR algorithm. All of the most parsimonious trees (MPTs) found in the first round were subjected to a second round of TBR to find all MPTs. Consistency (CI) and retention (RI) indices were calculated. Branch support is measured using Bremer support, jackknife, and bootstrap resamplings. Jackknife and bootstrap resamplings were obtained after 1000 replicates, and the values are shown as the difference in the frequency of a clade and the frequency when it is contradicted (GC values of Goloboff et al., 2003). If more than one MPT was found, a strict consensus was calculated. If the strict consensus was highly collapsed, the script written by Pol and Escapa (2009) was run in TNT to identify the rogue taxa and the precise causes (e.g., conflict of characters, lack of information, or both) for this behavior. To explore the evolution of certain characters, we mapped those characters using the TNT option of Optimize/Characters/Common mapping. Character numbers as in TNT (i.e., starting from 0). In Mesquite, character numbers are given as TNT number C1 (i.e., starting from 1). SYSTEMATIC PALEONTOLOGY TESTUDINES Batsch, 1788, sensu Joyce, Parham, andGauthier, 2004 TACUAREMBEMYS, gen. nov Etymology-After Mrs. Selva Kuster, who found the specimen.
Holotype-MGT-1185, which consists of an external partial mold of carapace that preserves the morphology of some scales: vertebrals 1-3 and the anterior part of vertebral 4; marginals 1-3; one pair of supernumerary anterior scutes; and medial portion of pleurals 1-3. It also preserves the morphology of some plates: the nuchal, peripherals 1-4 (partially), neurals 1-6, and costals 1-5. Holotype also includes two partial costal plates, one neural plate, and several shell bone fragments. Positive and negative casts of the holotype are deposited at Facultad de Ciencias, Uruguay, as FC-DPV-2761.
Diagnosis-Tacuarembemys kusterae differs from Condorchelys antiqua Sterli, 2008, andChubutemys copelloi Gaffney, Rich, Vickers-Rich, Constantine, Vacca, andKool, 2007, in the smaller size of the shell, type of ornamentation, having narrow vertebral scales, and having the sulcus between vertebrals 3 and 4 on neural 5. It also differs from Chubutemys copelloi in the absence of preneural bone and in the presence of anterior supernumerary scales. It shares with Sinemydidae, crown Pleurodira, and crown Cryptodira the presence of vertebrals narrower than pleural scales and sulcus between vertebrals 3 and 4 placed on neural 5. It shares with Araripemys barretoi Price, 1973, andLaganemys tenerensis Sereno andElShafie, 2013, the absence of cervical scale and the presence of a deep nuchal notch, allowing vertebral 1 reaching the anterior rim of the carapace; it differs from those taxa, however, in the absence of highly ornamented shells, contact between neural 2 and costal 3, the presence of anterior supernumerary scales, and the shape of costal 1. It also differs from A. barretoi in the shape of the nuchal and from L. tenerensis in the shape of peripherals. It is different from Prochelidella cerrobarcinae in the presence of nuchal notch, absence of cervical scale, presence of anterior supernumerary scale, and shape of vertebral 1. It differs from Sinemys lens Wiman, 1930, in the presence of marginal sulci on the peripherals only (at least in the anterior region of the carapace) and in the shape of the nuchal.
Autapomorphic characters of Tacuarembemys kusterae include carapace vaulted but not strongly convex, with a wide and well-developed nuchal notch (length/width D 0.30); a pair of triangular anterior supernumerary scales placed between marginal 1, vertebral 1, and pleural 1; wide nuchal plate with curved anterior and posterior borders; and narrow neural plates, with neurals 1 and 3-6 hexagonal and neural 2 rectangular in outline.

DESCRIPTION
The shell is represented by an external mold preserved in finegrained sandstone, which reflects the morphology of the major part of the carapace (ca. two-thirds of it; see Fig. 2), as well as several disarticulated, mostly fragmentary carapace bones. Some of the most complete bones, such as costal 2 and neural 4 ( Fig. 3), match the correlated regions of the mold. The estimated total shell length (straight sagittal length) is 180 mm (additional measurements in Table 1). The mold preserves the limits of scales and plates, but not the subtle shell surface ornamentation. This can be seen on the external surface of the bone fragments using a stereoscopic lens or a microscope (see below). The portion of the carapace mold posterior to neural plate 6 and the external borders posterior to peripheral 2 are not preserved (Fig. 2). Contrary to other continental Jurassic turtles (e.g., Kayentachelys aprix, Eileanchelys waldmani, Heckerochelys romani, xinjiangchelyids, sinemydids; Gaffney et al., 1987;Sukhanov, 2000Sukhanov, , 2006Matzke et al., 2004;Anquetin et al., 2009), the anterior border of Tacuarembemys kusterae is incised by a pronounced and deep nuchal notch. The deep nuchal notch resembles the one observed in Mongolochelys efremovi from the Upper Cretaceous of Mongolia and Naomichelys speciosa from the mid-Upper Cretaceous of North America (Khozatsky, 1997;Hirayama et al., 2000).
Macroscopically, the plates appear devoid of ornamentation (Figs. 3, 4A), except for small ridges that develop near the limits of the plates and are perpendicular to them. The external surface of a costal plate fragment was examined and small, randomly distributed pits are observed both microscopically ( Fig. 4A) and using the SEM (Fig. 5). In thin-section, small, randomly distributed vascular canals were observed (Fig. 6). Although the general morphology of T. kusterae resembles that of N. speciosa, the lack of any remarkable ornamentation in the former differs notably from the pustulate ornamentation found in solemydids (Lapparent de Broin and Murelaga, 1999;Joyce et al., 2011).

Plates
The nuchal plate is arch-shaped and wider rather than long. Its anterior margin is concave (due to a deep nuchal notch) and its posterior margin convex. It articulates with the first neural plate, the first pair of costal plates, and the first pair of peripheral plates. Half of its lateral sutures are not clear, and perhaps they were approximately coincident with the lateral borders of the first vertebral and first marginal scales.
The neural series is preserved from neural 1 to neural 6. Neural 1 is hexagonal and contacts with the nuchal, costals 1 and 2, and neural 2. Neural 2 is the smallest of the preserved neurals and is almost quadrangular, being slightly longer than wide. It contacts neurals 1 and 3 and costal 2 only. Neurals 3-6 are hexagonal. All are in contact with the previous and following neural and two costals (the corresponding costal and the anterior). Neurals 1 and 3-6 are subequal in size, with neural 3 slightly longer than the remaining neurals. The morphology and contacts of the neural series of T. kusterae is similar to those of Siamochelys peninsularis, a Middle Jurassic turtle from Thailand (Tong et al., 2002), but differ from those of stem turtles and N. speciosa (Gaffney et al., 1987;Hirayama et al., 2000).
The first pair of costal plates is wing-shaped and contacts peripherals 1-4. Costals 2-6 are roughly rectangular and much wider than long. The contacts with the neurals are described above. Due to the preservation of the specimen, contacts among costals 2-6 with the peripherals are missing. The presence or absence of costoperipheral fontanelles cannot be determined. Whereas most of the peripheral plates are quite poorly preserved, the anterior ones (1-4) can be reconstructed on the basis of some sutures and the anterior morphology of the first costal plate (Fig. 2).
Peripheral 1 is part of the nuchal notch. It contacts with the nuchal, costal 1, and peripheral 2. Peripheral 2 contacts with peripherals 1 and 3 and costal 1, whereas peripheral 3 contacts with peripherals 2 and 4 and costal 1. The shell bone fragments (thickness up to 5.25 mm) represent mostly carapace remains, including the left costal 2 and the neural 4, whereas a few others probably belong to the bridge region of the plastron.

Scales
The cervical scale is absent in T. kusterae, which seems to be due to the deep nuchal notch and the short marginal scales. The cervical scale has been lost independently in different clades of turtles (e.g., pelomedusoids and testudinids). However, in a few (e.g., the pleurodiran turtle Araripemys barretoi from Aptian-Albian of Brazil; Meylan, 1996) the absence of the cervical scale correlates with the presence of a deep nuchal notch. In both pelomedusoids and testudinids, the absence of the cervical scale allows medial contact between the first marginals. This medial contact is absent in T. kusterae and A. barretoi, where vertebral 1 reaches the anterior border of the carapace at the level of the nuchal notch.
Vertebral 1 is trapezoid and covers the major part of the nuchal plate, the first half of neural 1, and part of costal 1. Vertebral 1 is the narrowest of the preserved series. Vertebrals 2-4 of T. kusterae are as wide as the pleurals (as in S. peninsularis and crown Testudines), differing from the wide vertebrals of stem Testudines (e.g., Kayentachelys aprix, Condorchelys antiqua, and Indochelys spatulata) and N. speciosa (Gaffney et al., 1987;Datta et al., 2000;Hirayama et al., 2000;Sterli, 2008). Vertebral 2 is larger than the first, hexagonal, and broader than long. It covers the posterior half of neurals 1 and 2, the anterior part of neural 3, and parts of costals 1-3. Vertebral 3 is similar in shape and size to the preceding vertebral; it includes posterior half of neurals 3 and 4, the anterior half of neural 5, and parts of costals 4-6. The contact between vertebrals 1 and 2, vertebrals 2 and 3, and vertebrals 3 and 4 is located in the first, third, and fifth neural plates, respectively. The presence of the sulcus between vertebrals 3 and 4 and crossing neural 5 is also present in N. speciosa and crown Testudines. In some stem Testudines (e.g., K. aprix, I. spatulata, C. antiqua, and Eileanchelys waldmani), this sulcus crosses neural 6 (Gaffney et al., 1987;Datta et al., 2000;Sterli, 2008;Anquetin et al., 2009).
The anterior marginal scales are entirely included in the peripheral bones. Marginal scale 1 is small and roughly rectangular in shape, making contact with vertebral 1, the anterior supernumerary scale, and marginal scale 2. Although the cervical scale is absent, marginal scales do not contact each other in the midline, probably due to the deep nuchal notch present in T. kusterae. The left (right in the mold) marginal 1 and part of the marginals 2 and 3 are preserved, whereas the remaining marginal

Plate Histology
Thin-sections of a costal plate allowed the assessment of the histological structure of T. kusterae (Figs. 4B, 6). The plates show a thick layer of cancellous bone framed by external and internal cortices of compact bone.
The external cortex is thick (thickness about 1.60 mm; Figs. 4B, 6) and well vascularized with primary vascular canals, which are roughly perpendicular to the plate surface and seldom branching. The external end of these vascular canals could be related to the pits observed in the SEM (Fig. 5). Conspicuous growth marks are not observed, although there may be a few faint ones. Possible Sharpey's fibers can be observed (Fig. 6).  The cancellous bone is thick (thickness about 1.68 mm; Figs. 4B, 6) and strongly vascularized. Primary osteons with lamellae and bone cell lacunae are concentrically arranged and widespread.
The internal cortex is thin (thickness about 0.88 mm; Figs. 4B, 6) and vascularized with scattered single primary vascular canals, and it is composed of parallel-fibered bone. Bone cell lacunae are arranged parallel to the plate surface. The boundary between the cancellous bone and the internal cortex is sharply demarcated.

PHYLOGENETIC RESULTS
After two rounds of TBR, 240 most parsimonious trees (MPTs) of length 905 steps were found (Fig. S1), with a CI of 0.335 and a RI of 0.763. A list of synapomorphies common to all MPTs is shown in Appendix S2. The strict consensus of all MPTs  shows a poorly resolved topology at the base of Testudines, but in all MPTs T. kusterae is recovered inside Testudines (Figs. S1, S2). A big polytomy could be caused by the absence of information or by a wildcard taxon (see below) jumping between different positions in the MPTs. Consequently, we ran the script written by Pol and Escapa (2009), which identified T. kusterae as the most unstable taxon, jumping between two positions: (1) as the sister taxon of crown Pleurodira and (2) as the sister clade of Sinemys lens (see reduced consensus in Fig. 7). Following the script, contradictory characters do not cause the alternate position of T. kusterae and the scoring of several characters will help to solve its phylogenetic position (see Fig. S3).

Phylogenetic Position of Tacuarembemys kusterae
Although Tacuarembemys kusterae behaves as a wildcard taxon, the preserved characters allow us to place it inside Testudines. Testudines is supported in this analysis by two synapomophies common to all MPTs: only one suprapygal and small first thoracic rib, which are characters 136(0) and 200(2), respectively. These two characters are not preserved in T. kusterae, but other scored characters preclude T. kusterae from being a basal stem Testudines. The presence of the sulcus between vertebrals 3 and 4 on neural 5 (character 142) is a synapomorphy of the clade crownward to Chenyungchelys, whereas the presence of narrow vertebral scales (character 141) is a synapomorphy of the clade crownward to Siamochelys peninsularis. In stem Testudines (e.g., Kayentachelys aprix, Condorchelys antiqua, and Mongolochelys efremovi), vertebral scales are wider than pleural scales and the sulcus between vertebrals 3 and 4 is located on neural 6. The absence of shell sculpturing (character 123) positions T. kusterae in the clade crownward to Plesiochelys etalloni. Although only 7.5% (18 of 241 characters) of the characters are known for T. kusterae, those characters support referral to Testudines. More specimens will hopefully bring more information, in order to stabilize its phylogenetic position among Testudines and thus permit a better understanding of the evolution of this clade in Gondwana.

Character Evolution
MGT-1185 shows a combination of characters that are present in a variety of turtles, but there is no other turtle that exhibits all of them together, which justifies the recognition of a new taxon. The characters discussed below are characters that have been used for systematic purposes and that we consider worth discussing.
Ornamentation (Character 123)-The T. kusterae material described does not show the diagnostic characters that would allow it to be reliably included in a specific clade within Testudines. Although there are two characters shared with Naomichelys speciosa, the ornamentation of the latter consists of "raised tubercles that easily dislocate from the surface" (Joyce et al., 2011:83), which is markedly different from the small pits observed in T. kusterae. Scale sulci are of different shape as well (Hirayama et al., 2000: fig. 3). In our cladistic analysis, the absence of evident shell sculpturing is synapomophy of the clade crownward from Plesiochelys etalloni (see Fig. S5).
Absence of Cervical Scale (Character 137)-The absence of a cervical scale is a characteristic of T. kusterae, but it is not the only species in which this element has been lost (see Fig. S6). Although the cervical scale has been lost independently in several clades of turtles (e.g., pelomedusoids, testudinids, and Sinemys sp.), the morphology resulting from that loss is not the same in all cases. In testudinids and crown pelomedusoids, the cervical scale is absent but the anterior region of the carapace is not notched, resulting in the medial contact of both first marginals. In contrast, in T. kusterae, A. barretoi (Meylan, 1996;Fig. 8H), and Sinemys sp. (Brinkman and Peng, 1993), the anterior border is notched, the cervical is missing, but marginals 1 make no contact at the midline.
The presence of only one scale on one side of the shell is regarded as a shield abnormality (Zangerl, 1969;Pritchard, 2007). On the contrary, the presence of paired (bilaterally symmetric) anterior supernumerary scales seems to be relatively constant in some taxa (e.g., baenids) and seems not to be caused by shield abnormalities. The presence/absence of paired anterior supernumerary scales has been included in the present cladistic analysis, which shows (see Fig. S7) that the presence of anterior supernumerary scales appeared independently several times during turtle evolution. Nuchal notch length/nuchal notch width measured from Meylan and Gaffney (1991), Lapparent de Broin and Murelaga (1999), Hirayama et al. (2000), Sukhanov (2000), Milner (2004) Vertebral/Pleural Scale Relationships (Character 141)-The presence of vertebral scales 2-4 that are wider than pleural scales is well known to be plesiomorphic for Testudines because it is present in basal turtles such as Proganochelys quenstedti, K. aprix, I. spatulata, and Condorchelys antiqua (Gaffney et al., 1987;Gaffney, 1990;Datta et al., 2000;Hirayama et al., 2000;Sterli, 2008). Contrary to the condition present in basal turtles, T. kusterae has narrow vertebral scales (condition present crownward to Chengyunchelys; see Fig. S8).

Paleoecology
Histology can help infer the degree of aquatic adaptation in recent and fossil turtles (e.g., Ernst et al., 2006;Scheyer, 2007). Four categories were recognized by Scheyer (2007): I (terrestrial turtles), II (semiaquatic and mainly aquatic), III (fully aquatic), and IV (extreme adaptation to marine/aquatic environments). Plates of Tacuarembemys kusterae show a diploe structure (external cortex, cancellous bone, and internal cortex) that is indicative of categories I and II. The internal cortex is reduced in thickness, as in categories II and III. However, it represents half the thickness of the external cortex; i.e., disparity is not as great as in turtles of categories III and IV. Thus, this turtle is more likely to belong to category II (semiaquatic and mainly aquatic). Further evidence for the aquatic lifestyle is that the shell is dorsoventrally low; i.e., not strongly convex (e.g., see Sukhanov, 2000). All the preceding evidence agrees with the paleoenvironment proposed for the Tacuaremb o Formation: permanent and temporary streams and lakes (Perea et al., 2001(Perea et al., , 2009). However, doubts have been cast concerning the reliability of several means traditionally used to infer extinct turtle paleoecology (Benson et al., 2011).
Given the absence of cranial remains, nothing can be inferred about diet. If T. kusterae was durophagous, potential prey would have included some of the aquatic invertebrates (e.g., bivalves and gastropods) also found in the Batov ı Member (see Mart ınez et al., 1993;Mart ınez and Figueiras, 1993;Shen et al., 2004). Abundant mollusk molds have been found close to the type locality.
Late Jurassic, several (extant and extinct) clades originated and diversified (e.g., Cryptodira, Pleurodira, and Meiolaniformes; Sterli et al., 2012). However, Upper Jurassic continental deposits with vertebrates are scarce in Gondwana, and particularly in South America (Fig. 9A). For all of the above-mentioned reasons, this new turtle from Uruguay, though fragmentary, preserves valuable information from a key time in turtle evolution.
The components of the continental Upper Jurassic and Lower Cretaceous record of turtles in Gondwana and Laurasia are notably different from one another. The Upper Jurassic record of continental turtles of Gondwana is virtually nonexistent, because the Cañad on Asfalto Formation (where Condorchelys antiqua Sterli, 2008, was found) has recently been radiometrically dated as Early-Middle Jurassic (C uneo and Bowring, 2010). Fragmentary turtle remains from the Upper Jurassic of Ethiopia (labeled 'Testudinata indet.' in Fig. 9A) have been reported by Goodwin et al. (1999). If the proposed affinities with pelomedusids, plesiochelyids, and/or solemydids (termed 'pleurosternids' by Goodwin et al., 1999) is confirmed, it would provide interesting paleobiogeographic implications. On the contrary, a large number of Early Cretaceous taxa have been described from Gondwana. The oldest records of identified Cretaceous continental turtles from Gondwana come from the Barremian of Brazil (Gallo et al., 2009) and Morocco (Lapparent de Broin, 2000), the Aptian-Albian of Argentina (de la Fuente et al., 2007(de la Fuente et al., , 2011Gaffney et al., 2007), and the Albian of Australia (Gaffney et al., 1998;Smith, 2010;Smith and Kear, 2013). Early Cretaceous chelonian faunas from Africa were dominated by panpelomedusoid pleurodiran turtles (Lapparent de Broin, 2000), Recent taxa of which are restricted to South America and Africa (Bonin et al., 2006). In contrast, the Early Cretaceous chelonians from continental South America and Australia were dominated by the extinct meiolaniforms (Gaffney et al., 2007;Smith and Kear, 2013) and by panchelid pleurodiran turtles (de la Fuente et al., 2007(de la Fuente et al., , 2011Smith, 2010), of which the latter are now restricted to South America and Australasia (Bonin et al., 2006). The list of Early Cretaceous South American turtles is completed with the Aptian-Albian Santana fossils. However, the paleoenvironment of the Araripe basin is still debated due to the existence of both freshwater and saline deposits (Gaffney et al., 2006).
The Late Jurassic and Early Cretaceous continental turtle fauna from Laurasia was very different from that found in Gondwana (Fig. 9). Contrary to Gondwana, in Laurasia there is no evidence of the presence of crown Pleurodira and/or Meiolaniformes in the Upper Jurassic to Early Cretaceous. Besides, most of the components of the Early Cretaceous continental turtle fauna from Laurasia were already present from the Late Jurassic. That was not the case in Gondwana, possibly because of the scarcity of Upper Jurassic continental deposits. Continental environments from North America and Europe were dominated by pleurosternids, 'macrobaenids,' and solemydids (Gaffney, 1979;Brinkman et al., 2000;Joyce et al., 2011;P erez-Garc ıa and Ortega, 2011). A different scenario occurred in Asia, where one of the most diverse records of Late Jurassic to Early Cretaceous turtles can be found (Fig. 9).
The Asian record was dominated by 'xinjiangchelyids,' 'sinemydids,' sinochelyids, pantrionychians, pantestudinoids, and the stem-testudine Sichuanchelys sp. (Sukhanov, 2000, and references therein;Danilov and Parham, 2007;Rabi et al., 2010, and references therein). In the present cladistic analysis, T. kusterae behaves as a wildcard taxon, and one of its alternative positions is as a sister clade of Sinemys lens. We are cautious in recognizing this position, and we prefer to wait for more complete specimens before regarding T. kusterae as a sinemydid.