Introduction

The Rhynchocephalia is an ancient group of reptiles that originated in the early Mesozoic. Currently this group has low diversity, being represented by a single species, the famous ‘living fossil’ Sphenodon punctatus (Gray, 1842) from New Zealand (Jones et al., 2013; Cree, 2014; Herrera-Flores et al., 2017). In contrast to their current low diversity, Mesozoic rhynchocephalians were diverse, showing varied morphologies and a wide geographical distribution (Jones, 2006a, 2009; Rauhut et al., 2012; Martínez et al., 2013; Herrera-Flores et al., 2017). Among the earliest rhynchocephalians, species of the genus Clevosaurus Swinton, 1939 were the most diverse and widely distributed in the early Mesozoic. Clevosaurus hudsoni Swinton, 1939 was the first described species of the genus; it was named after F. G. Hudson, who discovered the fossil remains at Cromhall Quarry, England (Fraser, 1988). Since the description of C. hudsoni, nine species of Clevosaurus have been erected—C. bairdi Sues, Shubin, and Olsen, 1994, C. brasiliensis Bonaparte and Sues, 2006, C. convallis Säilä, 2005, C. latidens Fraser, 1993, C. mcgilli Wu, 1994, C. minor Fraser, 1988, C. petilus Young, 1982, C. sectumsemper Klein et al., 2015, and C. wangi Wu, 1994—and new records have been reported from localities in Belgium, Brazil, Canada, China, Great Britain, Mexico, and South Africa (Fraser, 1988, 1993; Wu, 1994; Sues et al., 1994; Duffin, 1995; Sues and Reisz, 1995; Säilä, 2005; Bonaparte and Sues, 2006; Reynoso and Cruz, 2014; Klein et al., 2015).

The anatomy of Clevosaurus is well known and the monograph of Fraser (1988) offers a very thorough review of the general morphology of this genus. It is recognized that the genus Clevosaurus is highly diverse, but the taxonomic validity of some Clevosaurus species has been questioned (Jones, 2006a). Hsiou et al. (2015) presented a review of C. brasiliensis that included a phylogenetic analysis of almost all known Clevosaurus species. Their study demonstrated that some species might not be valid taxa or are perhaps not directly referable to this genus. One of these conflicting taxa is C. latidens, a species described by Fraser (1993) from the Late Triassic fissure deposits of Cromhall Quarry, England. The uncertain taxonomic affinity of C. latidens and its dubious relationship with Clevosaurus have been noted in many previous studies (Jones, 2006a, 2009; Martínez et al., 2013; Hsiou et al., 2015; Klein et al., 2015), and some phylogenetic analyses even suggested a closer relationship with opisthodontians, but no taxonomic revision of this taxon has been carried out.

For a long time, the relationships among rhynchocephalians were poorly known, and most taxa were assessed by overall morphological similarities. The first phylogenetic analysis of the group was performed by Fraser and Benton (1989), followed by many different analyses, including new descriptions or redescriptions of taxa (e.g., Wu, 1994; Reynoso, 1996, 1997, 2000, 2005; Reynoso and Clark, 1998; Apesteguía and Novas, 2003; Rauhut et al., 2012; Martínez et al., 2013; Apesteguía and Carballido, 2014; Apesteguía et al., 2012, 2014; Cau et al., 2014; Hsiou et al., 2015). So far, all phylogenetic studies of the Rhynchocephalia have only used parsimony analysis, recovering a few distinct clades.More recently, Bayesian inference methods have been employed for phylogenetic analyses based on morphological characters (e.g., Parry et al., 2016; Wright, 2017), and recent studies suggest that Bayesian methods outperform parsimony for morphological data (O'Reilly et al., 2016; Puttick et al., 2017), recovering more accurate, but less precise results.

To clarify the doubtful taxonomic affinity of Clevosaurus latidens, we re-examined the type specimens and other material described by Fraser (1993). We updated the character matrix of a recent phylogenetic analysis of the Rhynchocephalia (Hsiou et al., 2015), recoded morphological characters for C. latidens, and performed both parsimony and Bayesian analyses. Our results confirm that C. latidens is not related to Clevosaurus, but represents a new genus. Our phylogenetic analyses recover similar topologies using both parsimony and Bayesian approaches. We employ the new phylogeny to propose formal names for two higher clades within Rhynchocephalia.

Material and methods

We re-examined the type material and other material described by Fraser (1993) as Clevosaurus latidens. All specimens of C. latidens consist of fragments of dentary, maxilla, and premaxilla housed in the collections of the Virginia Museum of Natural History and the University of Aberdeen. For anatomical comparisons, we reviewed several specimens of Clevosaurus from the paleontological collections of the University of Bristol and the University Museum of Zoology in Cambridge.

To explore the phylogenetic relationships of rhynchocephalians and the position of Clevosaurus latidens, we used the largest and most up-to-date data matrix of Rhynchocephalia (Hsiou et al., 2015). We added three taxa—C. sectumsemper Klein et al., 2015, Derasmosaurus pietraroiae Barbera and Macuglia, 1988, and Priosphenodon minimus Apesteguía and Carballido, 2014—and recoded some characters for C. latidens and Pelecymala robustus Fraser, 1986 after examination of the type specimens. The new matrix comprises 47 operational taxonomic units scored for 74 characters. We rooted the trees with the lepidosauromorph Sophineta cracoviensis Evans and Borsuk-Bialynicka, 2009. Two squamates, the Late Jurassic—Early Cretaceous Eichstaettisaurus Kuhn, 1958 and the extant Pristidactylus Gray, 1845, were also used as outgroups.

The revised taxon-character data matrix was analyzed using both equally weighted maximum parsimony and Bayesian inference. Parsimony analysis was performed in TNT v. 1.5 (Goloboff et al., 2008; Goloboff and Catalano, 2016), first using the ‘New Technology’ search options. The initial tree search used multiple replications with sectorial searches, four rounds of tree fusing, 10 rounds of drifting, and 200 ratcheting iterations. Following this, the generated most parsimonious trees (MPTs) were analyzed using traditional tree bisection and reconnection branch swapping. All recovered MPTs were then summarized in a 50% majority rule consensus tree, and clade robustness was assessed with Bremer decay indices (Bremer, 1994). Bayesian inference trees were estimated using MrBayes v. 3.2 (Huelsenbeck and Ronquist, 2001; Ronquist et al., 2012). The standard Mk model (Lewis, 2001) with gamma distribution priors for site rate variation was specified. The analysis was performed with four runs of four chains, run for 108 generations, sampling parameters every 1000 generations. The first 25% of sampled trees were discarded as burn-in. Convergence was assessed based on effective sample size (ESS) values >200. Results from the Bayesian analysis were summarized using a 50% majority consensus tree, revealing clades that have posterior probability values of ≥⃒ 50%. The data matrix and analytical scripts are included in the Supplementary Data Set.

Repositories and institutional abbreviations.—AUP= University of Aberdeen Paleontology Collection; BRSUG= Bristol University, School of Earth Sciences Collection; NMS=National Museums Scotland; SAMK=South African Museum; UMZC=University Museum of Zoology, Cambridge; VMNH=Virginia Museum of Natural History.

Systematic paleontology

Figure 1.

Fraserosphenodon latidens n. comb.; all specimens shown in labial, lingual, and occlusal views: (1–3) VMNH 524, holotype, left maxilla; (4-6) VMNH 525, paratype, right dentary; (7–9) VMNH 526, paratype, left maxilla; (10–12) VMNH 527, paratype, right maxilla; (13–15) VMNH 528, paratype, left maxilla. Scale bars=5mm (1–3, 7–9, 10–12); 3.5mm (4–6, 13–15).

Figure 2.

Fraserosphenodon latidens n. comb.: (1, 2) AUP 11191, right premaxilla, shown in labial (1) and lingual (2) views; (3–5) AUP 11192, right dentary, shown in labial (3), lingual (4), and occlusal (5) views. Scale bars=6mm (1, 2); 3.5mm (3–5).

Phylogenetic analyses

The parsimony analysis found 7176 MPTs of 265 steps, and the 50% majority rule consensus tree shows good resolution for most clades (Fig. 3.1). The consistency (CI) and retention indices (RI) for the 50% majority rule consensus tree are 0.38628 and 0.66403, respectively. No clade had a Bremer support score > 1 (complete statistics and associated files for both phylogenetic analyses can be found in the Supplemental Data). Generally, our results agree with those of other recent studies (Rauhut et al., 2012; Martínez et al., 2013; Apesteguía et al., 2014; Cau et al., 2014; Hsiou et al., 2015). One of the major differences is that our analysis recovered Pleurosauridae as the sister group of Sphenodontidae. The terrestrial Pamizinsaurus Reynoso, 1997 is the earliest diverging taxon within the Sphenodontidae, which includes two major clades. The first clade includes Ankylosphenodon Reynoso, 2000, Derasmosaurus Barbera and Macuglia, 1988, Oenosaurus Rauhut et al., 2012, and Zapatadon Reynoso and Clark, 1998 in a polytomy, whereas the second clade is well resolved, recovering the Early Jurassic Cynosphenodon Reynoso, 1996 and the modern Sphenodon Gray, 1831 as successive sister taxa to the clade comprising Theretairus Simpson, 1926 and Sphenovipera Reynoso, 2005. The strict consensus tree of the second analysis of Cau et al. (2014) also found Derasmosaurus, Oenosaurus, and Zapatadon in a similar polytomy, and forming the sister group of the clade comprising Sphenodon, Cynosphenodon, Sphenovipera, Kawasphenodon Apesteguía, 2005, and Theretairus. The close relationship of Sphenovipera and Theretairus has been constantly recovered in previous analyses (e.g., Martínez et al., 2013; Apesteguía et al., 2014; Hsiou et al., 2015).

Within clevosaurs, Brachyrhinodon Huene, 1910 was recovered as the earliest diverging taxon. All Clevosaurus species are grouped in a polytomy, which obscures the relationships between the species. The results for clevosaurs are quite similar to those recovered by the strict consensus tree of Hsiou et al. (2015). The only difference is that in their analysis, Polysphenodon Jaekel, 1911 appears as the earliest diverging taxon within Clevosauridae, but all other taxa were recovered in a polytomy. A similar polytomy for clevosaurs was also shown in the strict consensus tree of Rauhut et al. (2012). Our results agree with the work of Martínez et al. (2013) and Hsiou et al. (2015) in recovering Fraserosphenodon latidens as an early diverging opisthodontian. Indeed, we recovered F. latidens as the earliest diverging taxon within Opisthodontia. This clearly confirms that F. latidens is not referable to the genus Clevosaurus, and supports the erection of a new opisthodontian genus, as previously suggested (Jones, 2006a, 2009; Martínez et al., 2013; Hsiou et al., 2015; Klein et al., 2015). Within Opisthodontia, the relationships of eilenodontines are quite well resolved; our results only differ from the works of Martínez et al. (2013) and Cau et al. (2014) in finding Ankylosphenodon outside of Opisthodontia.

Another major difference compared to the previous analyses of Martínez et al. (2013) and Hsiou et al. (2015) is that the Triassic taxon Pelecymala was no longer recovered as closely related to Opisthodontia, but was found in a polytomy with early-diverging rhynchocephalians such as Rebbanasaurus Evans, Prasad, and Manhas, 2001, the clade of Sphenocondor Apesteguía, Gomez, and Rougier, 2012 and Godavarisaurus Evans, Prasad, and Manhas, 2001, and the clade Eusphenodontia.

Figure 3.

Consensus trees recovered from the phylogenetic analyses: (1) maximum parsimony 50% majority rule consensus tree; CI= 0.38628, RI=0.66403; node labels denote the proportion of MPTs that recover that node; (2) 50% majority rule consensus tree from the Bayesian-inference analysis, with clade credibility values (decimal proportions) labeled on the nodes.

Overall, the results of the Bayesian analysis (Fig. 3.2) resemble those of the parsimony analysis, but with considerably less resolution. Several large polytomies were recovered, but where clades are resolved, the clade credibility values are often moderately high. The Bayesian 50% majority rule consensus tree also recovered Pelecymala in a polytomy with early diverging rhynchocephalians, which confirms that this taxon is not related to opisthodontians as previously assumed (Martínez et al., 2013; Hsiou et al., 2015). The Bayesian tree did not recover clevosaurs as a monophyletic group; all of them were recovered in a large polytomy that obscures the relationships between the taxa. Relationships among other, later-diverging rhynchocephalians are unclear; many of them are part of a polytomy that includes Fraserosphenodon, but no clevosaurs. This result confirms that Fraserosphenodon is not closely related to Clevosaurus.

It should be noted that the Bayesian tree recovered a close relationship between the extant Sphenodon and the Jurassic Cynosphenodon, a close relationship between Theretairus and Sphenovipera, and pleurosaurs as a monophyletic group. The Bayesian tree did not recover Opisthodontia as a monophyletic group, but completely agrees with the parsimony tree for the interrelationships of eilenodontines, which are quite robust and well resolved.

Discussion

Among Mesozoic rhynchocephalians, clevosaurs were one of the most diverse groups. Clevosaurs are represented by three genera: Polysphenodon, Brachyrhinodon, and Clevosaurus. The first two genera are monospecific, whereas Clevosaurus currently has nine formally recognized species. The high diversity of the genus Clevosaurus, however, is debatable because of the doubtful validity/referral of some of the species, especially those based on poorly preserved or very fragmentary material, such as the three Chinese species (C. mcgilli, C. petilus, and C. wangi) or ‘C. latidens’ from Great Britain. According to Jones (2006a), the Chinese specimens are too poorly preserved to diagnose them as three distinct species, but clearly all of them belong to Clevosaurus. In contrast to the Chinese specimens, the referral of ‘C. latidens’ to Clevosaurus has been widely questioned (Jones, 2006a, 2009; Martínez et al., 2013; Hsiou et al., 2015; Klein et al., 2015).

Before the description of ‘Clevosaurus latidens,’ specimen AUP 11192, a dentary fragment, was tentatively related to Pelecymala based on its transversely wide teeth (Fraser, 1986). When Fraser (1993) formally described ‘C. latidens,’ he noted that the tooth morphology of the new ‘Clevosaurus’ species was quite similar to that of other taxa with transversely wide teeth such as P. robustus and Toxolophosaurus cloudi Olson, 1960 (Fraser, 1993). Some of the diagnostic characters of the genus Clevosaurus based on features of the skull could not be observed in ‘C. latidens’ for obvious reasons. However, at least the dentition of ‘C. latidens’ did not match that of Clevosaurus, which consists of larger, blade-like teeth with lateral flanges. It has been suggested that the tooth morphology of Clevosaurus was very specialized for a possible omnivorous or carnivorous diet (Jones 2006b, 2009; Rauhut et al., 2012; Martínez et al., 2013), whereas the dentary and maxillary teeth ‘C. latidens’ were more like those of herbivorous taxa. Fraser (1993) also pointed out that the wear facets on the teeth of ‘C. latidens’ suggested a propalinal movement of the lower jaw, which contrasts with the orthal jaw movement seen in Clevosaurus.

Based on dentary, maxillary, and premaxillary tooth morphology, as well as the suggested propalinal movement of the lower jaw, our review of ‘C. latidens’ specimens confirms that this taxon is not referable to Clevosaurus. Our phylogenetic analyses, including both parsimony and Bayesian approaches, confirm its position outside Clevosaurus. We rename ‘C. latidens’ as Fraserosphenodon latidens n. comb. The parsimony tree (Fig. 3.1) suggests that F. latidens is an earlydiverging opisthodontian, but not closely related to Pelecymala as was previously suggested by Fraser (1986, 1993), Martínez et al. (2013), and Hsiou et al. (2015). While reviewing the type specimens of Pelecymala (AUP 11140, 11214-11215), we noticed that the teeth of Pelecymala are not transversely broadened as had been described by Fraser (1986); in contrast, their shape is more conical, slightly curved, and labiolingually flattened. The tooth morphology of Pelecymala is actually more similar to that of some of the earliest diverging rhynchocephalians, which is also confirmed by our phylogenetic analyses (Fig. 3). A complete taxonomic redescription of Pelecymala appears necessary, but is beyond the scope of this study. The Bayesian tree (Fig. 3.2) could not recover the exact relationships of F. latidens, because this taxon is found in a polytomy that includes many other species. Like the parsimony analysis, however, the Bayesian approach recovered F. latidens as a genus distinct from Clevosaurus and not closely related to clevosaurs. Following the parsimony analysis, we consider F. latidens as an early diverging opisthodontian.

The parsimony analysis of Rhynchocephalia showed better resolution than the Bayesian approach. This result is not unexpected, because studies have shown that Bayesian methods are more accurate but less precise than parsimony-based analyses (O'Reilly et al., 2016). There are some minor differences between the internal branches in both trees, but several higher clades were recognized by both phylogenetic methods (Fig. 3). Some of these higher clades within Rhynchocephalia have been frequently recovered in other recent phylogenetic analyses, and have been informally named as ‘crown-sphenodontians,’ ‘derived-sphenodontians,’ or ‘eupropalinals’ (e.g., Apesteguía et al., 2012, 2014; Apesteguía and Carballido, 2014).

We propose formal names for two well-supported clades: Eusphenodontia and Neosphenodontia (Fig. 3). We define Eusphenodontia as the least inclusive clade containing Polysphenodon muelleri Jaekel, 1911, Clevosaurus hudsoni, and Sphenodon punctatus. In the 50% majority rule consensus tree, three unambiguous character transitions were recovered for Eusphenodontia under both ACCTRAN and DELTRAN optimization: (1) wear facets on marginal teeth of the dentary and/or on marginal teeth of the maxilla are clearly visible (character 46: 0 to 1), (2) the premaxillary teeth are merged into a chisel-like structure (character 49: 0 to 1), and (3) the palatine teeth are reduced to a single tooth row with an additional isolated tooth (character 52: 0 to 1). Neosphenodontia is defined as the most inclusive clade containing S. punctatus but not C. hudsoni. In the 50% majority rule consensus tree, Neosphenodontia is supported by the following six unambiguous character changes that are recovered under both ACCTRAN and DELTRAN optimization: (1) the relative length of the antorbital region is increased, reaching one-quarter to one-third of the complete skull length (character 1: 2 to 1), (2) the posterior edge of the parietal is only slightly incurved inward (character 18: 0 to 1), (3) the parietal foramen is found at the same level or anterior of the anterior border of the supratemporal fenestra (character 19: 0 to 1), (4) the palatine teeth are further reduced to a single lateral row (character 52: 1 to 2), (5) the number of pterygoid tooth rows is reduced to one or none (character 55: 1 to 2), and (6) the ischium is characterized by a prominent process on its posterior border (character 60: 1 to 2). The families Homoeosauridae, Pleurosauridae, and Sphenodontidae form in our analyses, as in others, the content of the clade Neosphenodontia. Levels of homoplasy in Eusphenodontia and Neosphenodontia are generally high, with individual character consistency indices (CI) often <0.5. For both clades, no individual character has a CI of 1 in the 50% majority rule consensus tree (for the complete list of characters, apomorphies, and other tree statistics, see the Supplemental Data). We consider the formal naming of these higher clades necessary to facilitate future discussion about the phylogenetic relationships of rhynchocephalians.

Conclusion

This study confirms previous doubts about the referral of ‘C. latidens’ to Clevosaurus. The recognition of ‘C. latidens’ belonging to a new genus now formally named Fraserosphenodon emphasizes the high generic diversity of Rhynchocephalia in the Mesozoic, especially among herbivorous taxa. Furthermore, our study demonstrates that the use of Bayesian approaches can be useful to contrast and validate phylogenies that were previously based only on parsimony methods. Bayesian inference exhibits generally lower resolution in some parts of the tree, but a few higher clades are strongly supported and are consistently recovered by both Bayesian and parsimony analyses.