The anatomy, paleobiology, and evolutionary relationships of the largest extinct side-necked turtle

Authors: E.-A. Cadena, T. M. Scheyer, J. D. Carrillo-Briceño, R. Sánchez, O. A Aguilera-Socorro, A. Vanegas, M. Pardo, D. M. Hansen, M. R. Sánchez-Villagra

Abstract

The largest extinct turtle from South America had males with horned shells.

INTRODUCTION

Since the extinction of non-avian dinosaurs, the northern Neotropics have harbored now-extinct vertebrates that have been at the extreme of large size within their respective clades (1). Among them are the largest snake (2), caimanine crocodile (3), gharial (4), and some of the largest rodents (5). One of the most iconic of these species is the gigantic turtle Stupendemys geographicus, as it is the largest nonmarine turtle ever known from a complete shell (until now rivaled only by the extinct marine turtle Archelon ischyros from the Late Cretaceous). It was first described in 1976 from the Urumaco Formation in northwestern Venezuela (6), but our knowledge of this animal has been based on partial specimens that have resulted in a problematic taxonomy, especially due to a lack of specimens with associated skull and shell elements. The species diversity of the giant turtles inhabiting northern South America during the Miocene is thus unclear (7, 8), with several forms having been the postcrania-based S. geographicus from the late Miocene, Urumaco region (6, 9–12); the skull-based Caninemys tridentata from the late Miocene, Acre region, Brazil (8); the controversial postcrania-based “S.” souzai, also from the late Miocene of Acre, Brazil (8, 13), currently attributed as Podocnemididae incertae sedis (14); and the skull-based Podocnemis bassleri from the late Miocene, Acre region (Loretto), Peru (15). The fossil record of large-sized littoral-freshwater Podocnemididae turtles of South America also includes the skull-based Carbonemys cofrinii, and the shell-based Pelomedusoides indet., from the middle to late Paleocene of Colombia (12).

We here describe several new shells and the first lower jaw specimens from discoveries made during regular fieldwork in the Urumaco region since 1994 (10, 16) and recent finds from La Tatacoa Desert in Colombia. Together, these fossils shed new light on the biology, past distribution, and phylogenetic position of giant neotropical turtles. First, we report a new size record for the largest known complete turtle shell. Second, our findings support the existence of a sole giant erymnochelyin taxon, S. geographicus, with an extensive geographical distribution in what were the Pebas and Acre systems (pan-Amazonia during the middle Miocene to late Miocene in northern South America). Third, we hypothesize that S. geographicus exhibited sexual dimorphism in shell morphology, with horns in males and hornless females.

RESULTS

Systematic paleontology

Testudines Batsch, 1788.

Pleurodira Cope, 1864 sensu Joyce et al., 2004.

Podocnemididae Cope, 1868.

Eymnochelyinae sensu Ferreira et al., 2018.

Stupendemys geographicus Wood, 1976.

Synonymy

Caninemys tridentata (8)

Stupendemys souzai (7, 13)

Stupendemys sp. (17)

Podocnemididae indet. (17)

Holotype

MCNC-244, medial portion of the carapace with associated left femur, fragments of scapulocoracoid and a cervical 8? (6).

Hypodigms

Specimens described in Wood (6): MCZ(P)-4376, much of the carapace, fragments of plastron, cervical 7?, both scapula-coracoids and a caudal vertebra; MCNC-245, a plastron lacking the epiplastra and entoplastron, two nearly complete costals, several peripherals, and one neural; MCZ(P)-4378, a right humerus. Specimen described as C. tridentata (8): DNPM-MCT-1496-R, nearly complete skull (Fig. 4, A to D). Specimens referred to as S. souzai (13): UFAC-1764, incomplete right humerus; UFAC-1163, cervical vertebra; UFAC-1294, left peripheral 1; UFAC-1544, left costal 2; UFAC-1547, right xiphiplastron; UFAC-1553, cervical vertebra; UFAC-1554, cervical vertebra; UFAC-4370, pelvic girdle; UFAC-5275, cervical vertebra; UFAC-5508, anterior margin of the carapace and left hypoplastron, and LACM-131946, nuchal bone, originally attributed to Stupendemys sp. (17). Specimens referred to as Podocnemididae ind. (17): LACM-141498, left lower jaw ramus, and Stupendemys sp. (17): LACM-138028, right scapula. New specimens described CIAAP-2002-01 (allotype), nearly complete carapace (Fig. 1, A to E); AMU-CURS-85, nearly complete carapace, left humerus, and right scapula-coracoid (Figs. 2, B and C, and 3, A to D); AMU-CURS-1098, plastron and anterior portion of carapace (Fig. 2, D and E); MPV-0001, nearly complete carapace and complete plastron (Fig. 2, F to M); OL-1820, left humerus (Fig. 3, E to H); AMU-CURS-233, fragment of femur (Fig. 3, I to P); AMU-CURS-706, lower jaw (see fig. S6); VPPLT-979, lower jaw (Fig. 4, E to L).

Range and distribution

Middle to late Miocene, Tatacoa Desert, Villavieja, Departamento del Huila, Colombia; late Miocene, Urumaco, Falcón State, Venezuela; late Miocene, Acre region, Brazil; Loretto region, Peru (Fig. 5).

Diagnosis

S. geographicus is recognized as a pleurodire based on (i) sutural articulation of pelvis with shell, (ii) loss of medial contact of mesoplastra, (iii) well-developed anal notch, (iv) fusion of gulars, (v) formed central articulations of cervical vertebrae, (vi) a well-developed processus trochlearis pterygoidei, and (vii) quadrate-basioccipital contact. It is a podocnemidid based on (i) a fully developed, medially extensive cavum pterygoidei with a completely developed pterygoid flange; (ii) an incisura columellae auris enclosing stapes and eustachian tube; (iii) an exoccipital-quadrate contact absent; and (iv) a cervical centra with saddle-shaped posterior condyles. It shares with Peltocephalus dumerilianus and Erymnochelys madagascariensis (i) a long parietal-quadratojugal contact; (iii) large postorbital bones; (iii) cheek emargination potentially reduced or absent; (iv) potentially advanced posterior roofing of the skull (reduced temporal emargination); (v) an articular with a processus retroarticularis posteroventromedially projected, differing from the ventrally projected Podocnemis spp. (Fig. 4, M to T, and fig. S6) acute tip of dentary at symphysis; and (vii) foramen chorda tympani enclosed in processus retroarticularis.

Further description and dimensions

Detailed anatomical descriptions, comparisons, and measurements of the fossilized bones and body mass estimation for S. geographicus and other fossil and extant giant turtles are presented in Fig. 3 (Q to V) and in the Supplementary Materials (text, figs. S1 to S6, tables S1 and S2, and data files S1 and S2).

Remarks

Skull: Unique among podocnemidids (and all other pleurodires, the side-necked turtles) in having greatly inflated maxillae, each with a ventral, tooth-like process, which, together with a single process formed on the midline of the premaxillae, form a tridentate condition in the upper triturating surfaces. Lower Triturating surface deep, forming an oval concavity, deeper than in any known living or extinct podocnemidid, labial ridge curved anteriorly ending in acute tip; lingual ridge is a blunt margin forming an accessory ridge that increases in height and width anteriorly and runs as a narrow ridge at the medial symphysis; high coronoid process; large dorsal opening of fossa Meckelii; the fossa Meckelii fills the posterior end of the jaw to such an extent that the area articularis mandibularis forms part of the posterior margin, and the fossa opens posterolaterally next to the jaw articulation. Shell: Carapace ≥2 m straight midline length, carapace low-arched, with irregular nodular contours on external surface and deep median notch at front; anterior border of nuchal-peripheral bones thickened and moderately to strongly upturned; carapace with massive anterolateral horns slightly projected ventrally in forms attributed as male; carapace dorsal bone surface smooth to striated or slightly pitted; posterior peripheral bones moderately scalloped along margins; thickness of carapace relatively thin at the costals. Shell (plastron): Pectoral-abdominal sulcus very anterior to mesoplastra, reaching almost the hyoplastra lateral notch level. Neck: Cervical vertebrae (probably 7 and 8) with neural arches relatively high in relation to anteroposterior lengths of centra, and articular facets of postzygapophyses forming acute angle of less than 90°; cervical 8? neural arch with large horizontal plane, prezygapophyses directed perpendicularly, thin bladelike spine on anterior face of neural arch and no ventral keel on centrum. Humerus: Humerus squat, massive; deep bicipital fossa between lateral and medial articular facets on ventral surface; prominent ridge traversing ventral surface of shaft from medial process to distal end, terminating just above lateral condyle; medial condyle broadest at anterior end; medial and lateral condyles facing very ventrally; straight to slightly slender shaft and triangular in cross section than circular. Femur: Femur squat, massive, greatly flattened dorsoventrally; breadth of tibial condyle approximately one-third total length of bone. Scapula: A dorsal strongly bowed scapular process with a flattened flange projecting laterally from the main axis.

Phylogenetic analysis

The first analysis (all taxa separated) produced 156,070 most parsimonious trees [MPTs; length = 1154, consistency index (CI) = 0.326, and retention index (RI) = 0.749]. The strict consensus tree (fig. S7) shows the lower jaws from Acre, Urumaco, and La Tatacoa in polytomy at the base of the Stereogenyini clade, sensu Ferreira et al. (18), with the same position for C. tridentata and S. geographicus as presented in Ferreira et al. (18). The second analysis [C. tridentata + Acre jaw and S. geographicus + (Urumaco, La Tatacoa jaws)] produced 1157 MPT (length = 1157, CI = 0.325, and RI = 0.748). Here, the strict consensus tree (fig. S7) shows C. tridentata and S. geographicus forming a monophyletic clade inside the Erymnochelyinae clade sensu Ferreira et al. (18), suggesting them to be closely related or potentially the same taxon. We favor the latter monospecific scenario based on the following (i) the three lower jaws from Urumaco, La Tatacoa, and Acre resemble each other in all morphological aspects, varying only in size and in preservation; (ii) the lower jaws from Urumaco, La Tatacoa, and Acre were found in localities and/or formations where shell material of S. geographicus was also found; and (iii) as Meylan et al. (8) stated, “there is a higher probability that the lower jaw, LACM-141498, does belong to Caninemys, and they are sufficiently complementary to suggest that they are from closely related taxa.” This scenario receives additional support from the third phylogenetic analysis, which produced 36 MPTs (length = 1180, CI = 0.319, and RI = 0.748). The strict consensus tree (Fig. 5A) and the time-calibrated cladogram pruned to the South American Erymnochelyinae clade (Fig. 5B) show S. geographicus at the base, as sister taxon to all remaining erymnochelyin turtles. This position is in agreement with the hypothesis presented by Meylan et al. (8) for C. tridentata (now S. geographicus), based on a relatively different character-taxon matrix. Jointly considering all these lines of evidence, we hypothesize that the skull of C. tridentata and the lower jaws described here together correspond to the skull of S. geographicus. It is thus both telling and fitting that turtle expert Eugene Gaffney, when supervising the exhibit of the reconstructed skeleton of S. geographicus at the American Museum of Natural History in New York, provided the skull of Caninemys for that model.

Body size and body mass estimation

The measurements of the new specimens are given in fig. S2 and table S2. Of particular interest is the new S. geographicus specimen CIAAP-2002-01 that we describe here. With its 286 cm parasagittal straight carapace length, it is not only the largest known specimen for this taxon but also the largest turtle shell found to date, considering that the hitherto largest known specimen is the so-called Vienna specimen of the turtle A. ischyros (NHMW-1977/1902/0001) with a shell length of 220 cm (19). Among Asian trionychids, giant forms have been reported from the Eocene of Pakistan, some reaching up to 2 m in shell length (20). Badam (21) reported on giant tortoises from the Pliocene of India that, based on reconstructed shell fragments, may have been larger than 3 m in carapace length.

We estimated the body mass using the straight carapace length method (see data file S1) (22). For the largest specimen, CIAAP-2002-01, we obtained an estimate of 871 kg [compared to the 744 kg obtained by Iverson (22) for MCZ(P)-4376, previously the largest and most complete specimen]. However, in the case of S. geographicus, to compensate for the effect of the large nuchal embayment, calculating the body mass estimate as the average between estimations based on the carapace midline and parasagittal lengths likely yields a more precise body mass estimate. Doing this results in a body mass estimate of 1145 kg for the CIAAP-2002-01 specimen.

Bone histology

The thin section of AMU-CURS-233 (Fig. 3, L to P) reveals an overall dense microanatomy with a central medullary region completely filled by cancellous bone, surrounded by a transitional zone with regular formed smaller spaces, which leads into a compact, external cortex. Because of erosion of the femur surface, the external-most layers of the bone are visible only in a few places.

The cortical tissue is increasingly dense toward the outer bone surface (Fig. 3M). The deeper parts of the cortex show a dense Haversian bone (Fig. 3N), consisting mostly of longitudinally arranged or slightly angled secondary osteons. In the more surficial parts of the cortex, remodeling into dense Haversian bone is prominent, but remnants of primary parallel-fibered bone matrix with numerous longitudinally arranged primary osteons are still present. In these remnants, cell lacunae are more irregular or of a roundish shape. The cortex also does not reveal growth marks that could be counted, with the exception of a single spot in the external-most cortical fragment that splits off from the main section due to delamination processes and gypsum growth. In this outermost-cortical layer, a few closely spaced lines (five lines?), interpreted as lines of arrested growth (LAGs), form an outer circumferential layer.

The cancellous bone in the center of the section (Fig. 3O) consists of short bone trabeculae and few irregular larger intertrabecular spaces. The trabeculae are secondarily remodeled and consist of lamellar bone.

The transitional bone (Fig. 3P) does not have distinct margins but is a zone of decreasing size of individual extravascular spaces and increasing bone compactness. Vascularization of the tissue is found in the form of longitudinally arranged osteons and only few circumferentially oriented ones. Remodeling by secondary osteons is extensive so that only interstitial pockets of primary parallel-fibered bone tissue are discernible. The overall bone compactness is 0.873, with modeled values at the center of 0.543 and at the periphery of 0.97 (see data file S2).

DISCUSSION

A pleurodiran turtle with a horned shell

In vertebrates, different body parts have independently evolved into protruding structures that are associated with a wide variety of purposes, e.g., defense or attack, mating, display, communication, or thermoregulation. Some of the most remarkable of these structures include horns, antlers, spikes, spurs, plates, tail clubs, and tusks (23–25). In turtles, a notable example is the posterolateral horns of the skull of the extinct meiolaniids (26, 27). Most examples, though, are connected to their shell, covering a diversity of types. Knobby ridges can be found on the carapaces of the extant matamata Chelus fimbriatus (28) and the alligator snapping turtle Macrochelys temminckii (29) and the extinct stem turtle Proganochelys quenstedti possessed serrations along the posterior shell margin (30). Among other examples, the extant spiny turtle Heosemys spinosa has peripherals with marginal spines (31) that disappear ontogenetically, and many groups of testudinid tortoises have highly lobulated and protruded anterior and posterior peripherals or anterior plastron edges (32). Horn-like structures at the anterolateral margin of the carapace, such as those we report here for S. geographicus, have previously only been documented in the Cretaceous nanhsiungchelyid Anomalochelys angulata (see fig. S8) (33). For this medium-sized (~65-cm straight parasagittal carapace length) extinct terrestrial turtle, one interpretation of the horns’ purpose was proposed as the protection of a large skull.

This hypothesis may also apply for S. geographicus, considering that we here interpret the massive skull DNPM-MCT-1496-R as its head. This specimen was previously described as C. tridentata, and it had lower jaws, which in several morphological aspects resemble the lower jaw of the extant South American big-headed turtle P. dumerilianus (Fig. 5 and the Supplementary Materials), including an acute symphyseal tip. Another feature that supports the robustness of the head of S. geographicus is the posterolateral opening of the fossa Meckelii in the newly recovered lower jaws described here (AMU-CURS-706 and VPPLT-979), implying a large main adductor tendon and associated musculature (17).

The occurrence of deep grooves in the massive horns of all three new specimens of S. geographicus from Urumaco described here (Fig. 1, C and D, and the Supplementary Materials) indicates that they were true horns with a bony core covered by a keratinous sheath that was strongly attached via the grooves, similar to horns of extant artiodactyl bovid mammals (34), and has been argued for meiolanid horns (26).

Sexual dimorphism

If the horns were for protection, then why do several S. geographicus specimens lack horns? The anteroventrally facing orientation of the horns is a distinct feature, suggesting that potentially they were exclusively used not only for protection but also for combat. We therefore hypothesize that the horned shells from Venezuela described here represent males of S. geographicus and that the horns served the main purpose of weapons in male-male combat behaviors. This hypothesis is consistent with the occurrence of similar structures in males of other groups of vertebrates, for example, in artiodactyl mammals (23, 34). In addition, in snapping turtles (Chelydridae), some of the largest extant freshwater turtles, males that occupy overlapping areas often establish dominance through fights (35). The elongated and deep scar in the left horn of CIAAP-2002-01 (see the Supplementary Materials) could be interpreted as a mark resulting from combat between males. Many extant tortoises use their protruding epiplastral horns for combat, often with the goal of flipping the opponent (36–38).

The putative S. geographicus males would also have been larger than females (see table S1), a pattern similar to that documented in the closely related extant taxon P. dumerilianus, which exhibits a male-biased sexual size dimorphism (39). Other sexually dimorphic traits of the turtle shell, such as a xiphiplastral concavity in males, or a deeper anal notch in males than in females (40, 41), are not distinct in S. geographicus, at least from a comparison between the specimens AMU-CURS-1098 (attributed to a male) and MPV-0001 (attributed to a female).

Paleoecology

The climate and the productivity of the environment, habitat size, and predation-competition interactions are some of the factors usually considered as triggers or in favor of gigantism (42, 43). We hypothesize that in the case of S. geographicus, a combination of several factors favored the evolution of its large size.

Habitat size, both in terms of individuals (home ranges large enough to sustain giant body sizes) and in terms of populations (species distribution ranges that can sustain long-term viable populations), was surely a major determinant. During the Paleogene and until the late Miocene [~66 to 5 million years (Ma)], after the retreat of the dominant marine conditions of the Cretaceous, northern South America harbored the most extensive freshwater and littoral ecosystems in its geological history. The coverage reached a particular peak during the Miocene, with the development of a large wetland and lake system known as the Pebas system (44), which offered not only increased connectivity between habitats but also the opportunity for the diversification and migration of faunas, including turtles. It seems that the size of these wetland habitats in northern South America during the Miocene facilitated the occurrence of gigantism not only in turtles (this study) but also in several vertebrate lineages such as crocodylians (Fig. 6 and table S3) and rodents (3–5).

Predation interactions could have also been involved in the evolution of large body size in S. geographicus, as it shared its habitat with gigantic crocodylians, including Purussaurus spp. and Gryposuchus spp., which could reach up to 10 m or more in body length. There is direct evidence of interactions between S. geographicus and large South American crocodylians, in the form of bite marks in Colombian and Venezuelan specimens, and an isolated tooth attached on the ventral surface of the carapace in the CIAAP-2002-01 specimen (see fig. S3).

Climate, particularly warmer temperatures, could have been a potential factor favoring the evolution of large body size in Miocene South American reptiles. For example, this causal link has been inferred for the Paleocene fauna of Cerrejón, Colombia, which includes the largest snake ever, Titanoboa cerrejonensis (2), and the largest Paleogene pelomedusoid turtles and crocodylians (12, 45). Although less warm than the Paleocene and the Eocene, the Miocene was also an epoch with notable climatic events that could have affected the body size of neotropical animal species, for example, the warm middle Miocene climatic optimum (MMCO) (46, 47), the global cooling between ~15 and 13 Ma known as middle Miocene climatic transition (MMCT), and continuous decreasing of global temperature during the late Miocene (48). The time range so far known for S. geographicus (middle Miocene to late Miocene) (this study) indicates that this taxon overcame the MMCT event. It exhibited a gigantic (and potentially its maximum) size during global cooling times (late Miocene) (Fig. 6). The latter rules out a direct and rather unlikely simple effect of climate on gigantism in neotropical Miocene reptiles. Thermally imposed upper limits to body mass are more likely than a simple tracking of changing temperature in body size evolution (49). Unfortunately, the climatic conditions of terrestrial ecosystems during the Miocene in tropical South America are still poorly known, and better reconstructions of climatic conditions await information from geochemical analyses of paleosols and carbonate isotopes. In addition, for neotropical faunas in general and reptiles in particular, the considerable gap in the South American Eocene and Oligocene fossil record is a major obstacle to a clear understanding of the effect of these climatic events on body size trends through time. It is therefore currently impossible to track the evolutionary path of evolution of body size that started during the Paleocene in detail or to establish whether body sizes of late Eocene and Oligocene neotropical reptiles remained large or decreased due, in part, to other cooling events such as the late Eocene-Oligocene transition from “greenhouse” to “icehouse.” To test the existence of a passive or driven trend in body size evolution (50), better sampling of the neotropical fossil record is needed. Both internal or external factors could be associated with such trends (51), and discoveries such as that reported here provide the primary evidence with which to start to understand the range of possibilities in morphospace occupation.

Turtles are a particularly challenging group when it comes to the identification of potential causal correlates in body size evolution, given the “atypical patterns” in relation to latitude they show in body size and in geographic range (52), as opposed to major tendencies identified for other vertebrate groups.

Last, the phylogenetic framework is likely an additional important factor, given the association of biological attributes such as body size and physiology to clades. Teasing out the relative importance of physiological boundaries related to clades is currently equally limited by the Eocene and Oligocene gap in neotropical faunas. For example, the large body size of S. geographicus could be an inherited ancestral trait, rooted in the Paleocene forms from Cerrejón, Colombia [Carbonemys cofrinii and its potential shell, Pelomedusoides indet. (12)]. Our phylogenetic analysis (Fig. 5) supports the view that S. geographicus and Ca. cofrinii both belong to the Erymnochelyinae clade but not as closely related taxa. What is clear is that at least two separate clades inside Podocnemididae exhibited large body size during the Miocene: one including S. geographicus and another with P. bassleri (15) (15.7-cm skull length, potentially 2 m carapace length) in the line of Podocnemis group. In other turtle clades of the neotropics, this trend is represented by Chelonoidis sp. (1 m carapace length estimate) inside the terrestrial Testudinidae and Chelus colombianus (70 cm carapace length estimate) within the freshwater-inhabitant Chelidae (Fig. 6).

Paleogeography

Adding to the previously known records of S. geographicus from Urumaco and Acre (6–8, 13, 17), we here report the first occurrence of this taxon in the well-known fauna of La Venta, Tatacoa Desert. This notably expands the known distribution of S. geographicus, highlighting that it likely was a common taxon throughout the entire Pebas system, well adapted to both fluvial conditions (La Venta and Acre) and fluvial-littoral conditions (Urumaco) (Fig. 5C). It is likely that the changes in the configuration of the Pebas and the posterior Acre systems due to the uplifting of the Andes starting in the middle Miocene (ca. 12.5 Ma) (53) (Fig. 6) had a deep impact on the populations of S. geographicus, considerably reducing their habitat size and leading to its final extinction, probably during the early Pliocene.

Paleodiet

Taking into account the morphology of the massive skull elements (skull and lower jaws, Fig. 4D) of S. geographicus, Meylan et al. (8) interpreted this turtle as a pleurodiran snapping turtle, involving a vacuum feeding system and capable of capturing and holding prey of very large size, including fish, small caimanines, and snakes. In this questionable interpretation, it was a carnivore much like the extant cryptodires Macrochelys, Claudius, and Staurotypus, which also exhibit a depression in the upper triturating surface and have lower jaws with a well-developed symphyseal hook (8). The very acute symphysial end and wider anteromedial triturating surface of the well-preserved jaw (VPPLT 979 specimen) from La Tatacoa described here indicate that S. geographicus may have had a diet much broader than one consisting of the abovementioned vertebrate preys. It could have had a more diverse diet. For example, it could have had a generally durophagous diet, crushing hard-shelled prey such as mollusks with the help of its large triturating surface and facilitated by its large main adductor tendon and associated musculature. Increasing the diet niche breadth would have favored maintaining a very large body size in this turtle, resulting in a body size–environment productivity correspondence (42).

Another previously underestimated aspect of paleodiet is the potential of large extinct turtles having acted as seed dispersers for many plant species. A recent review of frugivory and seed dispersal in extant turtles (54) highlighted that many species consume fruits, and thus potentially disperse the seeds, even if fruits are not considered part of their standard diet. Seasonally, high-energy fruits from, e.g., palms (Araceae) can even form the major part of Amazonian turtles’ diets. This is the case for the closest extant relative of S. geographicus, the big-headed Amazon river turtle, P. dumerilianus, where (55) found that fruits and seeds formed the most diverse component of its stomach contents and that palm seeds were the most common item (55).

Because of its huge gape size, S. geographicus could have swallowed even the largest South American fruits and thus qualify as a megafaunal frugivore and seed disperser [sensu (56)]. In general, larger turtles also include more fruits in their diet than do smaller ones; for example, in the extant Asian big-headed turtle, Platysternon megacephalum, there is a positive relationship between body size and amount of fruit in their diet (57). Overall, S. geographicus could thus have been a highly efficient seed disperser [sensu (58)].

Paleohistology and life history considerations

As with the previously analyzed shell bones of S. geographicus (from CIAAP-2002-01) (59), our histological analysis of the femur did not reveal anything unusual about Stupendemys growth, only that it is overall comparable to the microanatomical build and the histology of smaller turtles. The high amount of Haversian bone in the femur fragment might be related to the giant size as pointed out by Foote (60) or by advanced age of a skeletally mature specimen, as is tentatively indicated by the tightly spaced LAGs in the outer circumferential layer. The estimated compactness values of AMU-CURS-233 are comparable to those of other aquatic, nonmarine turtles (61).

We see the almost universal conserved arrangement of scutes of turtles in the gigantic specimen described here, emphasizing how the developmental program of turtles (62) results in early differentiation in which prolonged growth does not result in changes in epidermal structures. S. geographicus probably lived for at least 110 years to be able to reach the largest recorded size we report here, assuming a growth rate similar to that of extant, large turtles (59).

MATERIALS AND METHODS

Institutional repositories

The fossils referred here are in the collections of American Museum of Natural History, New York, USA; Alcadía Bolivariana de Urumaco, Urumaco, Falcón State, Venezuela (AMU-CURS); Centro de Investigaciones Antropológicas, Arqueológicas y Palentológicas (CIAAP) of the Universidad Nacional Experimental Francisco de Miranda, Coro, Falcón State, Venezuela; Departmento Nacional de Produçaõ Mineral, Divisaõ de Geologia e Mineralogia, Ciências da Terra, Rio de Janeiro, Brazil (DNPM-MCT); The Geological Museum, Geology Survey Institute, Bandung, Indonesia (K); Natural History Museum of Los Angeles, Los Angeles, USA (LACM); Museo de Ciencias Naturales de Caracas, Caracas, Venezuela (MCNC); Museum of Comparative Zoology-Harvard University, Cambridge, USA [MCZ(P)]; Museo Paleontológico de Villavieja, Villavieja, Huila Department, Colombia (MPV); Naturhistorisches Museum Wien, Vienna, Austria (NHMW); Universidad Simón Bolívar, Caracas, Venezuela (OL; specimens housed in the Museo Paleontógico de Urumaco); and Museo de Historia Natural La Tatacoa, La Victoria, Huila Department, Colombia (VPPLT).

Phylogenetic analyses

To explore the phylogenetic position of S. geographicus, three separate maximum parsimony analyses were run using PAUP 4.0 (63) and using the character-taxon matrix of Ferreira et al. (18) as the original template with some modifications (see the Supplementary Materials). For all the analyses, Pr. quenstedti, Notoemys laticentralis, and Platychelys oberndorferi comprised the outgroup taxa; all the 245 characters were considered equally weighted, and multistate states were treated as polymorphic. Heuristic search, random search for 10,000 replicates, and tree-bisection reconnection option were performed, seed 1000, holding one tree per replicate and collapse branches if minimum length is zero. Strict consensus trees and their decay index (Bremer support) were also obtained. For the first analysis, we considered each of the three giant lower jaws from Acre (17), Urumaco, and La Venta (described here) as separate taxa, as well as C. tridentata (8) and S. geographicus, with the addition of information from previous and the new specimens described here. A second analysis considering the lower jaw LACM-141498 from Acre as belonging to C. tridentata as considered originally by Meylan et al. (8) and the lower jaws AMU-CURS-706 from Urumaco and VPPLT-979 from La Tatacoa as belonging to S. geographicus was performed. For the third analysis, we considered a single taxon, S. geographicus, formed by the new and previously described S. geographicus shells and postcrania; the three lower jaws from Acre, Urumaco, and La Tatacoa; and the skull of C. tridentata (see fig. S7). Twelve morphocline characters were treated as ordered characters (14, 18, 19, 71, 95, 96, 99, 101, 119, 129, 174, and 175) following Ferreira et al. (18). Results are also presented in a time-calibrated cladogram of South American Erymnochelyinae turtles (Fig. 5B) based on this and previous studies (12, 64).

Body mass estimation

Body mass estimation of S. geographicus and some other taxa mentioned in table S1 was obtained using the correspondence between carapace length and body mass reported by Iverson (22) in extant representative of all lineages of turtles. Specifically, we used the general allometric equation y = ax^b^, where y is the body mass (in grams), x is the carapace length (in centimeters), and a and b are the correlation coefficients established for each of the taxa (see the Supplementary Materials) (22). Considering that none of the taxa included in this study were part of Iverson’s study, we used the coefficients of the closest phylogenetic and/or similar lifestyle representative, for example, in the case of S. geographicus as it was also used by Iverson (22), we used the coefficients established for Podocnemis unifilis; for A. ischyros and D. coriacea (both marine turtles), we used the coefficients of Chelonia mydas; for Megalochelys sivalensis and Chelonoidis niger (both tortoises), we used the coefficients of Geochelone elegans; and for Rafetus swinhoei (freshwater soft-shelled turtle), we used the coeficientes of Apalone (Trionyx) spinifera.

Bone histology

We sectioned a shaft fragment of a femur of S. geographicus (AMU-CURS-233) recovered from a site next to the gas pipeline at El Mamón locality, Urumaco, Falcón state, Venezuela (11°13′1.46″N; 70°16′51.2″W). The shaft section was roughly oval shaped, with the longest axis of 8 cm and a perpendicular shorter axis of 6.2 cm. The bone was cut with an iron hand saw and processed afterward, following standard petrographic thin-sectioning procedures (65). The thin section was studied and analyzed using a compound microscope (DM 2500M, Leica) with a digital camera (DFC 420C, Leica). Comparative material of S. geographicus included already published shell bone sections (59), and overall bone compactness was calculated using Bone Profiler software (66).

Maximum size versus climate and geological events

We plotted the largest as-preserved or estimated length of skull, lower jaw, and/or carapace of turtles and crocodylians from each of the neotropical Neogene to Quaternary fossil sites, putting them in context with the global climatic curve of Zachos et al. (46) and the major geological and geographical events of northern South America. We included the following lineages of Erymnochelyinae, Podocnemidinae, Chelidae, and Testudinidae, and for the Alligatoridae, Gavialidae, and Crocodylidae (Fig. 6, fig. S9, and table S3), adding also the largest reported extant representatives. We excluded from this plot very recently immigrant lineages of Geoemydidae, Kinosternidae, Emydidae, and Chelydridae, and turtles that occasionally reached South America, for example, Trionychidae, as well as sea turtles and the extant Galápagos tortoises (gigantism due to phylogenetic history and island isolation). The extremely fragmentary Charactosuchus spp. were also excluded considering that they are still controversial if they are truly members of Crocodylidae (67).

Supplementary Material