A New Clevosaurid from the Triassic (Carnian) of Brazil and the Rise of Sphenodontians in Gondwana

The early evolution of lepidosaurs is marked by an extremely scarce fossil record during the Triassic. Importantly, most Triassic lepidosaur specimens are represented by disarticulated individuals from high energy accretion deposits in Laurasia, thus greatly hampering our understanding of the initial stages of lepidosaur evolution. Here, we describe the fragmentary remains of an associated skull and mandible of Clevosaurus hadroprodon sp. nov., a new taxon of sphenodontian lepidosaur from the Late Triassic (Carnian; 237–228 Mya) of Brazil. Referral to Sphenodontia is supported by the combined presence of a marginal dentition ankylosed to the apex of the dentary, maxilla, and premaxilla; the presence of ‘secondary bone’ at the bases of the marginal dentition; and a ventrally directed mental process at the symphysis of the dentary. Our phylogenetic analyses recover Clevosaurus hadroprodon as a clevosaurid, either in a polytomy with the Late Triassic to Early Jurassic Clevosaurus and Brachyrhinodon (under Bayesian inference), or nested among different species of Clevosaurus (under maximum parsimony). Clevosaurus hadroprodon represents the oldest known sphenodontian from Gondwana, and its clevosaurid relationships indicates that these sphenodontians achieved a widespread biogeographic distribution much earlier than previously thought.

Southern Brazil (Fig. 1); Santa Maria Formation (Santa Maria Supersequence, Candelária Sequence), Rosário do Sul Group, Paraná Basin; Carnian, Late Triassic 31,32 (Fig. 1a-c). Clevosaurus hadroprodon was recovered from immediately beneath the layers that contained the cranial and postcranial remains of the cynodonts Exaeretodon sp. and Trucidocynodon sp. and a distal portion of femur that closely resembles the early dinosaur cf. Pampadromaeus (Fig. 1b). The presence of these fossils referred to Exaeretodon, Trucidocynodon and a Pampadromaeus-like form places this locality within the Hyperodapedon AZ at the base of the Candelária Sequence. Recent high-precision U-Pb zircon geochronology data recovered a weighted mean 206 Pb/ 238 U date of 233.23 ± 0.73 Ma for typical Hyperodapedon AZ sites 33 , which can be biostratigraphically correlated to the sedimentary layers containing C. hadroprodon. As such we conservatively consider the age for the fossil material to be Carnian (Late Triassic). most Clevosaurus species the nasal processes tend to be long, with the condition in C. brasiliensis 19 being the most similar to C. hadroprodon in relative length. However, differing from C. hadroprodon, C. brasiliensis also possesses a posterodorsal process of the premaxilla 19,21,28 , which is typical of most other known sphenodontians (e.g., Homoeosaurus, Palaeopleurosaurus, Piocronus, Pleurosaurus, Priosphenodon) and is also present in many archosauromorphs as well as the lepidosauromorphan Kuehneosaurus 22 . The absence of this process evidenced that only maxilla could have formed the posterior margin of the external naris. This feature is also seen in the early rhynchocephalians, Gephyrosaurus 35 and Diphydontosaurus 34 , respectively, as well as in Sphenodon.
A single premaxillary tooth is also present in the Clevosaurus bairdi 18 , C. brasiliensis 25 as well as in Kallimodon (pers. obs. TRS) and Sphenodon. The single premaxillary tooth of C. hadroprodon is massive and cylindrical making it more characteristically tusk-like. The longitudinal groove on the premaxillary tooth is likely a shallow facet on the distal surface (terminology following Smith and Dodson 37 ) of the tooth to accommodate the tip of the corresponding tusk-like tooth in the first dentary tooth position.
Maxilla. The right maxilla of the holotype (MMACR PV-027-T) is exposed in lateral view (Fig. 2a,b), preserving the portion directly associated with the tooth row, but is otherwise badly crushed and missing the articular portions of the anterior and posterior margins. The preserved portion is mostly the robust, smooth "secondary bone" (see below) that forms at the bases of the teeth which is typical of many sphenodontians. A small portion of the facial process is exposed posteriorly and appears to form a curved posterior process rising above the terminal tooth. This may be the jugal process but is too poorly preserved to be certain. The left maxilla is embedded in the matrix deep to the other elements, but the anteriormost portion is visible in medial view inferior to the premaxilla. Anterior to the two preserved teeth (rotated slightly posteriorly due to breakage of the jaw) is what appears to be a short "diastema". Anterior to this is a deeper notch, presumably the articulation facet for the premaxilla. Although poorly preserved, the anterior margin of the facial process appears similar to that of most Clevosaurus species in being tall, vertical, and ascending from above the anterior margin of the premaxillary facet. The maxillary dentition is similar to that of the dentary (see below).
Dentary. The anterior portions of both the left (medial view) and right (lateral view) dentaries are present in MMACR PV-027-T (Fig. 2a,b) and the lateral view of the anterior portion of the left dentary is preserved in MMACR PV-028-T (Fig. 2c,d). All three of the dentary specimens are crushed making it difficult to evaluate the anatomy inferior and posterior to the more robust anterior alveolar portions. None of the lower jaw specimens preserve discernable postdentary elements. Based on the preserved portion (nearly 10 mm) of the tooth row in the left dentary of the holotype, this element is relatively long and deep. The best-preserved portion of both dentaries is the massive subdental bone associated with the anterior portion of the tooth row. On the holotype dentaries this subdental bone is exposed in both lateral and medial views. The lateral side is convex, and bears a distinct, ventrally directed, mental process (particularly apparent on MMACR PV-028-T, Fig. 2c,d). There appear to be several small mental foramina near the symphysis (holotype MMACR PV-027-T, Fig. 2a,b), but the surface is heavily pitted, and it is not possible to confidently distinguish the pittings from possible foramina. The medial surface is more vertical with a tall, shallow longitudinal groove extending from the base of the anteriormost tooth to approximately the eighth tooth position. The anteriormost end of this groove appears to extend to the symphysis, but it is not clear if this is not an artifact of the damage to the symphysial portion of the element. More posteriorly the subdental bone flattens, appears blade-like, and bears an articulation facet presumably for the squamous contact with the articular and possibly the prearticular. Inferior to the subdental bone is the deep Meckel's groove, which is widely open at the symphysis, narrows posteriorly as a short isthmus, and then expands again as it continues posteriorly. The inferior and posterior portions of the dentary are not sufficiently well preserved to further evaluate the Meckel's groove. The symphyseal portion of the dentary was formed from a relatively thin, ventrally oriented tab that projected below the inferior margin of the dentary.
In comparison with other sphenodontians, the dentary of Clevosaurus hadroprodon is unique in its overall structure. It is apparently similar in its substantial depth but with a much longer tooth row than those of adult C. brasiliensis (see Romo de Vivar and Soares 38 for a review of the ontogeny of C. brasiliensis mandibles). Many sphenodontians with relatively long tooth rows such as living Sphenodon and such fossil taxa as Sphenocondor, Sphenovipera, and Tingitana 29,39,40 also have mandibles that are long and gracile. Most crown-group sphenodontians are characterized by having dentaries that are relatively short and deep.
The shallow groove dorsal to the Meckelian canal near the symphysis is similar to that described for Cynosphenodon 41 , but differs in Clevosaurus hadroprodon in being longer and possessing a narrow, rather than deep, ridge of bone separating it from the Meckelian canal. If the narrow extension of this groove beneath the base of the tusk-like dentary tooth towards the symphysis is not an artifact of preservation, then, as with Cynosphenodon 41 , it may indicate that the holotype specimen of C. hadroprodon is a juvenile (probably between the T3-T4 stages of Robinson 42 ).
The Meckelian canal of Clevosaurus hadroprodon is most similar to that found in C. brasiliensis (Hsiou et al. 25 ) and Sphenotitan leyesi 43 in being widely open with a constriction just posterior to the symphysis. A dentary (NHMUK R6102) referred to Clevosaurus hudsoni also has a widely open Meckelian canal, but lacks a conspicuous isthmus posterior to the symphysis, but another dentary (UMZC T1307) for the same taxon does exhibit this morphology suggesting it may be variable (TRS, pers. obs.). The groove dorsal to the Meckelian canal in C. hadroprodon is more pronounced than the narrow groove found in one of the dentaries of C. hudsoni (UMZC T1307), but in the case of C. hadroprodon. the dorsal margin of the groove is bounded by bone rather than by the ventral edge of "secondary bone".
Dentition. The holotype has heterodont dentition in the upper and lower marginal tooth rows characterized by the distinct dental regions developed in mesiodistal (anteroposterior) sequence ( Fig. 2a-d). This includes one anteriormost successional tooth, represented by the single, large, tusk-like tooth on the dentary (or 'caniniform' successional tooth, see Apesteguía et al. 29 ), and an equivalent tusk-like tooth in the premaxilla. Clevosaurus hadroprodon has an alternating tooth series distal to the first tooth position in the dentary represented by at least three teeth. There could be additional dentary teeth in this series, but these are not observable in both dentaries due to overlapping of the jaw elements. The anterior portion of the maxilla also has an alternating tooth series, represented by the first six preserved maxillary teeth in the right maxilla. Finally, the alternating series in both jaws is succeeded distally by an additional tooth series, comprised of six preserved teeth in the right maxilla. The number of teeth in the additional series on the dentaries is unknown due to poor preservation of the posterior dentary region.
The single successional tooth of the premaxilla is massive, conical, blunt (tip may be worn), and vertically oriented. The base is firmly ankylosed to the premaxilla and there is no evidence of "secondary bone" (sensu Harrison 26,27 ; Jones 22 ; AKA "secondary dentine" of Fraser 28 ; since the histological characteristics of this tissue have not yet been demonstrated we have adopted the more inclusive term 'secondary bone' but use quotes to denote uncertainty of its specific structure, but see Jones 22 for a review of alternative terminology).
The tusk-like tooth in the dentary occupies the first tooth position above the symphysis. This tooth, similar to that of the premaxilla, is mesiodistally massive, conical in shape, and apicobasally procumbent (approximately 30°), and is attached to the tip of the dentary slightly below the attachment region of the remaining dentary teeth.
The remaining teeth of the dentaries and the maxillae are similar in morphology. The longest of the dentary tooth rows preserves approximately 19 tooth positions and the most complete maxillary tooth row has 12 tooth positions, though both are broken posteriorly suggesting that several teeth are missing in both the upper and lower marginal tooth rows. Apart from the tusk-like teeth in the first dentary tooth position and in the premaxilla, (2019) 9:11821 | https://doi.org/10.1038/s41598-019-48297-9 www.nature.com/scientificreports www.nature.com/scientificreports/ all the preserved tooth crowns are similar in being labiolingually compressed, generally triangular, and with distinct mesiodistal carinae, but lacking the labiolingual flanges common to other sphenodontians. The teeth also differ from each other in relative size along the tooth row with shorter, narrower (mesiodistally) teeth anteriorly in the dentary and maxilla and taller, wider teeth posteriorly, but with smaller teeth interspersed among these larger teeth. The exact pattern cannot be determined due to damage along each of the tooth rows. There is no evidence in any of the specimens of active tooth replacement.
In the dentaries there is a thin ridge of "secondary bone" that obscures the tooth-jaw contact in both labial and lingual views with the labial being more prominent. All tooth bases of the holotype are fused and ankylosed to the apex of the jaws and lack any discernable evidence of active tooth replacement. There is no clear evidence of this tissue at the bases of the maxillary teeth, but only the labial surface is clearly visible. The "secondary bone" lacks any evidence of wear facets. comparative osteology of Clevosaurus hadroprodon to other rhynchocephalians. Clevosaurus hadroprodon can be classified within the Rhynchocephalia, and more specifically within Sphenodontia, based on several characters of jaw morphology: both anterior and posterior tooth series apically placed and ankylosed to the jaw bone labial margin; presence of 'secondary bone' deposition along the tooth-jaw contact; a dorsoventrally deep and vertically oriented symphyseal margin of the dentary; deep ventral crest of the dentary medial wall; and the presence of an anterior canine tooth that is apically placed and fused to the dentary. Although some of these features also occur within acrodontan squamates (which generally have a remarkable degree of convergence with sphenodontians), the combination of features in C. hadroprodon is only seen in sphenodontians (see Supplementary Data 1).
Clevosaurus hadroprodon shares with most rhynchocephalians the presence of acrodont tooth implantation in association with 'secondary bone' . This does differ from the early rhynchocephalians Gephyrosaurus and Diphydontosaurus in that these two taxa lack 'secondary bone' and are characterized by pleurodont attachment of polyphyodont teeth along the entire tooth row (Gephyrosaurus), or, the combination of anteriorly pleurodont and posteriorly acrodont attachment (Diphydontosaurus) 21,28,34,35,44 . Clevosaurus hadroprodon is thus more similar to later evolving forms by having a fully acrodont tooth implantation, and most of them also develop the 'secondary bone' in mature individuals at the tooth-jaw contact following the dental line 22,28,29,44,45 . However, Clevosaurus hadroprodon has a less developed layer of 'secondary bone' . Whether this reduced development of 'secondary bone' is related to taxonomy or ontogeny is unknown (absence is considered a juvenile feature 46 ).
Among sphenodontians, Clevosaurus hadroprodon is additionally unusual in its tooth morphology. The teeth are labiolingually compressed, lacking the development of labiolingual features of the marginal dentition common to most other sphenodontians (e.g. a posterolingual crest in Clevosaurus species, Kallimodon, Homeosaurus, Pleurosaurus, and Palaeopleurosaurus-TRS, pers. obs.; broadly expanded posterior teeth with anteriorly directed concavity teeth in opisthodontians). The absence of labiolingual dental features also occurs in the heterodont teeth of Whitakersaurus from the Late Triassic of New Mexico, USA 47 , and Rebbanasaurus and Godavarisaurus, from the Middle Jurassic-Early Cretaceous of the Kota Formation, India 21 . Among the latter, Clevosaurus hadroprodon also shares with Rebbanasaurus an extremely deep symphysial margin of the dentary, although differing from the latter in lacking striated tooth crowns.
The presence of a single tooth on the premaxilla of Clevosaurus hadroprodon is shared with the sphenodontians Priosphenodon and Sphenodon as well as the clevosaurs C. brasiliensis 19,25 , C. bairdi 18 and specimens currently attributed to Clevosaurus sp. from China 17,22 , differently from the pattern observed in other clevosaurs, such as C. hudsoni, C. convalis and C. sectumsemper that possess 2-4 premaxillary teeth 11,28,48 . In the stem rhynchocephalians Gephyrosaurus and Diphydontosaurus there are typically five or more small pleurodont teeth on the premaxilla 34,35 , and Planocephalosaurus possesses four subpleurodont premaxillary teeth ("semi-pleurodont" 28,34 ).
The anterior tusk-like tooth in the premaxilla of Clevosaurus hadroprodon is most similar to C. brasiliensis in being tusk-like (e.g., Hsiou et al. 25 ; TRS pers. obs.), and not chisel-shaped as in Sphenodon, Priosphenodon, and Vadasaurus (composed of three fused premaxillary teeth 49 ). Additionally, an enlarged successional tooth ("canine-tooth") in the dentary has also been described for Cynosphenodon and Sphenovipera 23,40 , but these occupy tooth positions posterior to the first tooth. The most similar condition to the enlarged, tusk-like, anteriormost tooth of the dentary in C. hadroprodon is possibly the enlarged, albeit of a much smaller relative size, anteriormost dentary tooth of Sphenocondor. However, in Sphenocondor the symphysial portion of the dentary is poorly preserved making it difficult to confirm the exact dimensions of this tooth 27 . A single, large tooth in the first dentary tooth position is also present in Cynosphenodon huizachalensis 23,41 . However, in C. huizachalensis this tooth is conical and preceded by an edentulous gap in adults and in juveniles is preceded by at least three "anterior hatchling teeth" and one "first generation successional tooth" 41 . According to Apesteguía et al. 29 the precise homology between the 'caniniform' tooth and other successional teeth remains uncertain in fossil taxa, and in this way, we consider that the tusk-like, anteriormost tooth of the dentary in C. hadroprodon as a successional tooth that is larger than subsequently successional teeth 29 .
As noted above, although important similarities exist between Clevosaurus hadroprodon and Mexican and Indian forms (Cynosphenodon and Rebbanasaurus), the combination of features in C. hadroprodon is unique among known sphenodontians and other clevosaurs, thus supporting its designation as a new species of Clevosaurus.

Discussion
The labiolingually compressed marginal dentition of Clevosaurus hadroprodon is generally similar to the overall dimensions of the "cut and slice" sphenodontian tooth form 22 including the marginal teeth of Clevosaurus spp. and the sphenodontine sphenodontians 50 . However, Clevosaurus hadroprodon differs from these taxa (and nearly all other sphenodontians) in the lack of obvious wear facets on the teeth indicating that the feeding strategy of (2019) 9:11821 | https://doi.org/10.1038/s41598-019-48297-9 www.nature.com/scientificreports www.nature.com/scientificreports/ this new taxon likely did not utilize the intense oral food-processing common to most sphenodontians. The sharply pointed dentition is more similar to that of Agama 51 and juvenile Uromastyx 52 -though these acrodontan lizards also show heavy wear facets in the teeth, particularly in older individuals-and suggests a similar arthropod-based diet. It is possible that the specimens of C. hadroprodon represent juveniles that had not yet been able to achieve significant wear in the marginal dentition, but the presence of "secondary bone" indicates that these specimens had undergone some appreciable ontogenetic development (sensu Duffin 46 ). The presence of a single, large tooth in the premaxilla and the dentary of C. hadroprodon is similar to the one observed in Agama, but in C. hadroprodon these teeth are tusk-like (straight and blunt) rather than sharply pointed and recurved. It is possible that such teeth could be used to subdue a prey item, but it is also possible that they served to aid in non-feeding behaviors such as mate competition or defense. The lack of palatal elements precludes assessment of this dentition and associated implications on feeding strategy, mechanism, or diet.
The teeth of Clevosaurus hadroprodon represent the oldest occurrence of the typical fully acrodont dentition of sphenodontians anywhere in the world, thus providing insights on the early stages of the development of the acrodont dentition in sphenodontians. In sum, the dental morphology in C. hadroprodon demonstrates that the presence of a fully acrodont dentition, reduction in premaxillary tooth count and development of large successional teeth on the premaxillae, typical of most sphenodontians, had evolved by the Late Carnian. Conversely, the simple, triangular shape of the marginal teeth in C. hadroprodon suggests that accessory crests-a feature common to the dentition of most sphenodontians-possibly evolved after the acquisition of the fully acrodont dentition in sphenodontians. But, as stated above, this simple tooth form may be an ontogenetic feature and additional, more complete specimens are needed to verify either of these possibilities.
Phylogenetic analysis recovers Clevosaurus hadroprodon within Clevosauridae. In a parsimony analysis (Fig. 3a) C. hadroprodon resolves within the genus Clevosaurus, whereas in an analysis utilizing bayesian optimality criterion (Fig. 3b), C. hadroprodon is recovered in a polytomy comprised of Brachyrhinodon and the Clevosaurus species. The erection of a new species is justified, although it lacks significant support given: (1) the overall morphological differences between the new species compared to Clevosaurus spp. (see comparative osteology section); and (2) the result of the bayesian inference, which does not recover the genus Clevosaurus as monophyletic. The poor resolution among the clevosaurids and related taxa in the bayesian topology, in addition to the position of Polysphenodon in the parsimony analysis, demonstrates the need for careful and detailed anatomical and systematic revision of Clevosauridae to clarify the interrelationships within these sphenodontians (see also Hsiou et al. 25   www.nature.com/scientificreports www.nature.com/scientificreports/ Information). Clevosaurus hadroprodon also represents an early diversification of specialized sphenodontians that occurred within the Gondwanan region of Pangaea prior to the Late Triassic.
A major component of this early diversification of sphendontians is driven by the origins of clevosaurs, within which Clevosaurus hadroprodon figures among the oldest known representatives (Fig. 4), along with Brachyrhinodon and Polysphenodon from the late Carnian-Norian 55 of Britain (see also Supplementary Fig. 1 for maximum clade credibility tree and Supplementary Data 2).
The occurrence of Clevosauridae in the late Carnian is notable in that it is approximately 10 my younger than the oldest known rhynchocephalian, cf. Diphydontosaurus ("Vellberg jaws" specimens) from the Ladinian (Middle Triassic) of Germany 5 and only 18 million years younger than the estimated divergence between Gephyrosaurus and all other rhynchocephalians 2 . Therefore, clevosaurs comprised the earliest evolving major group of lepidosaurs (with six species that eventually dispersed throughout five different continents), and seem to have developed a widespread geographical distribution very early in their evolutionary history. By the end of the Carnian, clevosaurs were present in at least two distantly located regions of Pangaea that today constitute parts of the United Kingdom and Brazil 18,22,25 . By the Early Jurassic clevosaurs had further dispersed across the fragmenting remains of Pangaea into North America 18 , Asia 17 , and South Africa 16 . As such, clevosaurs represent the earliest group of lepidosauromorphs to achieve such cosmopolitan distribution.
The dentition of Clevosaurus hadroprodon represents the oldest occurrence of the typical fully acrodont dentition of sphenodontians anywhere in the world, thus providing insights on the early stages of the development of the acrodont dentition in sphenodontians. For instance, the morphology of Clevosaurus hadroprodon suggests that accessory crests that are commonly observed on the dentition of most sphenodontians developed at a later stage in sphenodontian evolution, after the acquisition of the fully acrodont dentition, but not to the exclusion of other dental specializations such as the tusk-like teeth present in this taxon. It further suggests that the reduction in the number of premaxillary teeth and development of large successional teeth on the premaxillae, also typical of most sphenodontians, occurred by the Late Carnian, along with the development of a fully acrodont dentition.
Clevosaurus hadroprodon (Fig. 5), along with Clevosaurus brasiliensis, indicate that some of the oldest known sphenodontians evolved in South America, and along with Indian taxa, such as Godavariasaurus and Rebbanasaurus, indicate that some of the earliest diverging members of the Sphenodontia occurred in geographically distant parts of southern Pangaea, during the initial breakup of Gondwana. This illustrates the importance of the role of the Gondwanan lepidosaur fauna in our growing understanding of the earliest stages of sphenodontian evolution and global biogeographic distribution.  (Fig. 1a-c). The Triassic exposure occurs in a ravine of about six meters depth composed of fine to medium-grained sandstones, intercalated with mudstone layers and conglomerates (Fig. 1b). Carbonate concretions occur along most of the section, and the fossils tend to occur in association with the concretions. The specimens referred to Clevosaurus hadroprodon were originally embedded in concretions, abundant at the lowermost conglomeratic level of the outcrop. Other specimens, representing other taxa, were collected in the fine sandstone levels, also associated with concretions. They include partial lower jaws and some postcranial elements (e.g., humerus) referred to the traversodontid cynodont Exaeretodon riograndensis, partial skull referred to the probainognathian cynodont Trucidocynodon sp., and fragments of long bones of a dinosaur specimen, among other indeterminate materials. A fragment of a femur resembles the early dinosaur Pampadromaeus barberenai 56,57 ; although further preparation and comparisons are needed to clarify its taxonomy. The uppermost portion of the exposure consists of pinkish non-fossiliferous massive sandstone (Fig. 1b). The presence of specimens referred to Exaeretodon, Trucidocynodon and a Pampadromaeus-like form allows correlation of this outcrop as part of the Hyperodapedon AZ, at the base of the Candelária Sequence. The fossil samples from this outcrop are relatively scarce in comparison to other Triassic localities in the state of Rio Grande do Sul and a precise estimation of its relative age is only tentative. Exaeretodon, Trucidocynodon and a Pampadromaeus-like form have also been recovered from the Janner Site, in the municipality of Agudo. At this locality these taxa are part of a more diverse fauna including probainognathian (e.g. Trucidocynodon riograndensis) and traversodontid (Exaeretodon riograndensis) cynodonts 58,59 , dinosaurs and rhynchosaurs. Based on its faunal content and its relationships with other outcrops from southern Brazil, the Linha Bernardino site can be also correlated to the Ischigualasto Formation (Ischigualasto-Villa Unión Basin, Argentina 60,61 ). A recent analysis provided a high-precision U-Pb zircon geochronology to some sections of south Brazil Triassic exposures, including a typical Hyperodapedon AZ locality (i.e. Cerro da Alemoa) 33 . The recovered weighted mean 206 Pb/ 238 U date of 233.23 ± 0.73 Ma is slightly older than the datum for the Ischigualasto Formation 33 . Nonetheless, the combined data suggest that the age of the faunal association of the Linha Bernardino site is more likely Late Carnian (MCL, pers. comm.).
Phylogenetic analyses and time scaled-tree. Taxon scoring changes. Sophineta was deleted from the dataset because only some of the skull and mandible elements can be confidentially linked to a single taxon, with all the referred postcranial material lacking any anatomical connection to the skull and mandible elements (TRS pers. obs.). Since deletion of all postcranial scorings would highly reduce the usefulness of this taxon as an outgroup by re-scoring several cells with missing data, we deleted this taxon altogether from the analysis.
All specimens attributed to Clevosaurus brasiliensis were merged into a single operational taxonomic unit (OTU), since all of them have been subject of a recent revision of this species and confidently attributed to a single Rogue taxon identification. An initial set of analyses (Analysis 1) was conducted with all taxa initially available in the data set. This analysis resulted in consensus trees with important sectors of the tree highly unresolved (results available in Supplementary Information). Therefore, we conducted a rogue taxon identification analysis using the RogueNaRok algorithm, which can improve tree accuracy and resolution when specific wildcards are the main factor contributing to poorly resolved nodes 62 . We also utilized the TNT pruning trees algorithm to detect wild card taxa in specific nodes of the consensus tree inferred with maximum parsimony. The results indicate that the removal of Clevosaurus convalis, Sphenovipera and Theretairus would highly improve phylogenetic resolution. A second set of analyses were conducted after removal of those three taxa, which yielded better trees that are reported in the main text. phylogenetic analyses. Maximum parsimony analysis was conducted in TNT v. 1.1 63 using 100 tree replicates obtained by random addition sequence (RAS), and searching for new tree topologies with tree bisection and reconnection (TBR), saving 100 trees per replication. For Analysis 1, a total of 907 most parsimonious trees (MPTs) were obtained with 258 steps each. After removal of the three rogue taxa, the 40 remaining taxa were analyzed in a new set of analyses (Analysis 2). In Analysis 2, we recovered a total of eight MPTs with 249 steps each.
The Bayesian inference analysis was conducted using Mr. Bayes v. 3.2.6 64 using the Mkv model for morphological data 65 , and with rate variation across characters sampled from a gamma distribution. Each analysis was performed with two independent runs of 1 × 10 7 generations each, with eight chains per run and four swaps attempted per swapping generation. The relative burn-in fraction was set to 50% and the chains were sampled every 50 generations. The temperature parameter for the four chains in each independent run was set to 0.02. Convergence of independent runs was assessed through the average standard deviation of split frequencies (ASDSF < 0.01) and potential scale reduction factors (PSRF ≈ 1 for all parameters) calculated at the end of the Bayesian runs. We used Tracer v. 1.6 66 to determine whether the runs reached stationary phase and to ensure that the effective sample size (ESS) for each parameter was greater than 200.
The strict consensus tree calculated from the eight MPTs from the parsimony Analysis 2 was time-scaled using the package strap 67 for the software R 68 . The range of time for each taxon was delimited consulting the relevant literature (see Supplementary Data 3 and 4). It was employed in the scaling, the Brussate parameter 69 , in which all branch lengths are shared equally among the tree, in order to avoid zero values (i.e. only retaining positive length for each branch).

Data Availability
The phylogenetic analyses, time scaled-tree and geological unit and time range references of the Rhynchocephalia taxa data of this study are available as Supplementary Information.